Research & Publications
Research & Publications
International Journal Papers
2026
Paper Title:
Fuzzy Superpixel Segmentation with Anisotropic Total Variation Regularization
Authors:
Ng, T. C., Choy, S. K., Tang, M. L., Regelskis, V., & Lam, S. Y.
Abstract:
This paper presents a superpixel segmentation algorithm that integrates anisotropic total variation regularization within a fuzzy clustering framework. While isotropic total variation is well-known for its edge-preserving properties, its non-adaptive nature often leads to over-regularization. In contrast, the anisotropic model formulates superpixel regularity in relation to image contours, thereby preventing the loss of image details in areas of high contour density during optimization. Compared to classical segmentation algorithms that employ non-adaptive regularization, the proposed content-adaptive approach enhances superpixel regularity while maintaining boundary adherence to image contours. Furthermore, to optimize the functional effectively, an alternating direction method of multipliers along with the enhanced Chambolle’s fast duality projection algorithm are employed. Competitive experiments against existing regular segmentation algorithms demonstrate that our proposed methodology achieves superior performance in terms of boundary recall, compactness, and shape regularity criteria, outperforming these methods by an average of at least 3%, 5%, and 3%, respectively. Furthermore, when compared with irregular segmentation algorithms, our approach achieves the best results in terms of compactness, contour density, and shape regularity criteria, with average improvements of at least 56%, 22%, and 45%, respectively.
»Cite this article: Ng, T. C., Choy, S. K., Tang, M. L., Regelskis, V., & Lam, S. Y. (2026). Fuzzy Superpixel Segmentation with Anisotropic Total Variation Regularization. Mathematics, 14(3), 404.
»DOI:10.3390/math14030404
2025
Paper Title:
Transforming cold chain logistics: a reversible vehicle routing approach for sustainable and efficient delivery of perishable goods
Authors:
Lam, H. Y., Ho, G. T. S., Mo, D. Y., & Wong, L.
Abstract:
As global demand for perishable goods rises, logistics providers face challenges in managing temperature-sensitive products. Traditional vehicle routing models often neglect cold chain requirements, including temperature control and strict delivery schedules. Additionally, sustainability concerns drive the need for eco-friendly logistics practices, as passive packaging methods contribute to waste. This study introduces a reversible cold chain vehicle routing model that integrates the Internet of Things and a multi-objective optimisation framework for real-time monitoring. Incorporating cold chain time windows, recycling constraints, and social sustainability, the model demonstrates how technology can enhance cold chain logistics and support circular economy goals.
»Cite this article: Lam, H. Y., Ho, G. T. S., Mo, D. Y., & Wong, L. (2025). Transforming cold chain logistics: a reversible vehicle routing approach for sustainable and efficient delivery of perishable goods. Enterprise Information Systems, 19(5-6), 2492737.
Paper Title:
Enhancing efficiency in the picker-to-parts E-commerce warehouse: a perspective based on order lifecycle and operational behavior computing
Authors:
Dong, H., Huang, M., Lam, H. Y., Mo, L., Zhuang, X., & Zuo, M.
Abstract:
Over 80% of e-commerce warehouses use the picker-to-parts model, where employees collect items within the warehouse. Despite its widespread use, research that integrates the order lifecycle and employee behavior is limited. This article addresses this gap by exploring operational behavior computing and human-machine integration to optimize picker-to-parts efficiency. First, we outline the order-planning workflow from an order lifecycle perspective. We then conduct a literature review to identify practical applications of academic insights. The article concludes by suggesting future research directions, such as developing online coordinated optimization techniques and integrating multiple technologies to address data quality and uncertainty.
»Cite this article: Dong, H., Huang, M., Lam, H. Y., Mo, L., Zhuang, X., & Zuo, M. (2025). Enhancing efficiency in the picker-to-parts E-commerce warehouse: a perspective based on order lifecycle and operational behavior computing. Enterprise Information Systems, 19(5-6), 2448832.
Paper Title:
Does “climate change” equal “global warming”? A corpus-based study of lexical choices related to climate change in three UK newspapers from 2018 to 2025
Authors:
Lam, Y. M., & Lam, B. S. Y.
Abstract:
Climate lexical choices are crucial for effective communication and understanding in climate change discussions. Variations in these lexical choices may lead to misunderstandings among scientists, journalists, and policymakers. To reduce potential ambiguities in climate communications, this study aims to provide empirical evidence of the usage patterns of five lexical choices related to climate change, namely, “climate change,” “climate crisis,” “climate emergency,” “global warming,” and “global heating,” in three UK newspapers spanning from 2018-2025. Using Python, we calculate the co-occurrence frequency of the selected lexical choices and construct confidence intervals, then investigate, quantitatively and qualitatively, when and how the lexical choices were used. We found that despite the different political leanings, no clear lexical choice preferences were identified among the three newspapers. Contrary to common perceptions, we discovered an unrecognised common usage pattern in “climate change” and “global warming”―the two are not interchangeable in their applications in the UK newspaper context. In providing this timely empirical assessment of the usage patterns of lexical choices related to climate change, we invite journalists, scientists, and policymakers to reflect on whether the lexical choices, particularly “climate change” and “global warming”, should be used interchangeably in the future.
»Cite this article: Lam, Y. M., & Lam, B. S. Y. (2025). Does “climate change” equal “global warming”? A corpus-based study of lexical choices related to climate change in three UK newspapers from 2018 to 2025. Humanities and Social Sciences Communications, 12(1), 1852.
Paper Title:
Client Association and Resource Allocation for NOMA-Enabled Hierarchical Federated Learning With Non-IID Data
Authors:
Li, T., & Song, L.
Abstract:
Hierarchical federated learning (HFL) provides a new client-edge-cloud paradigm to reduce the network burden brought by centralized FL. However, the current HFL framework mainly adopts orthogonal multiple access (OMA) techniques for client-edge communication, failing to cope with massive connections with limited bandwidth resources. Besides, the highly skewed data distribution across the clients can undermine the convergence, which leads to more rounds to reach the target performance. In this paper, we consider a non-orthogonal multiple access (NOMA)-enabled HFL framework that provides a spectrum-efficient approach to enable massive client connectivity. Under this framework, we investigate the joint client-edge association and resource allocation problem to optimize the dual-efficiency. We first reformulate the original problem as a bi-objective optimization problem using dynamic weights to balance the resource and data heterogeneity. After that, we formulate the edge-client association problem as a many-to-one matching game with externalities and swap the connected pairs until a stable matching is obtained. For each temporary edge-client association, we solve the CPU frequency and transmission power control problem for all clients using convex optimization methods. We conduct experiments under various unbalanced data distributions and communication environment settings. The results demonstrate that our method outperforms other benchmarks regarding client scheduling, transmission and resource allocation.
»Cite this article: Li, T., & Song, L. (2025). Client Association and Resource Allocation for NOMA-Enabled Hierarchical Federated Learning With Non-IID Data. IEEE Transactions on Network Science and Engineering.
Paper Title:
Deep reinforcement learning-based approach for dynamic routing in quick-commerce e-fulfilment systems
Authors:
Mo, D. Y., Tsang, Y. P., Lam, H. Y., & Chung, K. T.
Abstract:
This study investigates the deep reinforcement learning (DRL)-based approach to manage dynamic same-day delivery orders in e-fulfilment systems connected to physical and internet logistics network. To enable real-time response to ad-hoc changes, a DRL-based approach which employs the actor–critic mechanism is proposed to generate dynamic vehicle routing solutions for q-commerce order additions and cancellations. Computational experiments were conducted to compare the performance of OR-Tools (ORT) and the proposed method in static and dynamic environments. The experimental results showed that the proposed deep reinforcement learning q-fulfilment routing optimizer (DRLQRO) offered similar performance to ORT in a static environment with higher vehicle capacity, while it outperformed ORT in a dynamic environment with lower vehicle capacity, achieving cost savings of 12.3–19.55%. The DRLQRO features adaptability and robustness, making them suitable for e-fulfilment systems in the digital twin era.
»Cite this article: Mo, D. Y., Tsang, Y. P., Lam, H. Y., & Chung, K. T. (2025). Deep reinforcement learning-based approach for dynamic routing in quick-commerce e-fulfilment systems. International Journal of Logistics Research and Applications, 1-24.
Authors:
Liu, H., Ye, S., Ma, Y. T., Wang, Y., & Chan, T. T.
Abstract:
With the rapid advancement of robotics technologies, a group of robots are able to communicate with one another by wireless transmissions and form a robot swarm. Robot swarm has many applications and a typical one is target search in which swarm robots are sent to places that might be dangerous for human workers, and they coordinate with one another to search for targets such as survivors in a disaster. However, motion coordination of swarm robots for target search has received little attention especially when the targets are mobile. In this work, we develop a motion coordination algorithm for swarm robots to search for targets in an unknown area. Our basic idea is to divide the search area into grids and build a gray-scale map in which each grid is associated with a gray scale indicating the efficiency of searching targets in this grid. The Voronoi diagram is adopted to coordinate swarm robots to search different portions of the search area for maximizing search efficiency. By theoretical analysis, our motion coordination algorithm is validated to ensure that all static targets are guaranteed to be found. We derive an upper-bound on the total time for robots to traverse all the grids in the search area. Extensive simulations are conducted and the results show that the proposed motion coordination algorithm outperforms the state-of-the-art and achieves a success rate of over 90% in finding all mobile targets with low search latency. Note to Practitioners—This paper was motivated by the problem of target search in an unknown area (e.g., the search for Malaysia Airlines MH370). With advanced robotics technologies, swarm robots could be sent to the area to perform a search mission. This work suggests a motion coordination algorithm for swarm robots to search for both static and mobile targets in an unknown area with possible obstacles. The proposed algorithm divides the search area into grids and each grid is associated with a gray scale indicating the efficiency of se…
»Cite this article: Liu, H., Ye, S., Ma, Y. T., Wang, Y., & Chan, T. T. (2025). Motion coordination of swarm robots for mobile target search. IEEE Transactions on Automation Science and Engineering.
Authors:
Tang, X. P., Tian, Y. Z., Wang, Y., & Wu, C. H.
Abstract:
Count data is a type of data derived from the number of times an event occurs per unit of time, and zero-truncated count data refers to count data without zero. Modeling for this type of data primarily utilizes a single zero-truncated distribution, such as the zero-truncated Poisson (ZTP), zero-truncated Bell (ZTBell), and zero-truncated negative binomial (ZTNB) distribution. However, except for ZTNB model, fewer relevant models involving heterogeneous count data have been studied. Therefore, in this article, we propose a zero-truncated Bell-Poisson mixture (ZTBPM) regression model based on Bell and Poisson distributions and study the parameter estimation method of this model. Monte Carlo simulation verifies that the ZTBPM model fits better than the ZTP, ZTBell, and ZTNB regression models, providing more options for modeling zero-truncated count data and solving the heterogeneity problem more effectively. Finally, the ZTBPM regression model is applied to a set of outpatient data analysis to study the risk factors affecting the number of residents’ outpatient visits. This is of great significance for better understanding the health status and medical needs of residents, optimizing the allocation of medical resources, and guiding the formulation of medical policies.
»Cite this article: Tang, X. P., Tian, Y. Z., Wang, Y., & Wu, C. H. (2025). Zero-truncated Bell-Poisson mixture regression model and its application to outpatient data. Communications in Statistics-Simulation and Computation, 1-10.
Authors:
Luo, F., Jiang, P., Ho, G. T. S., & Zeng, W.
Abstract:
Multiobject tracking of ships is crucial for various applications, such as maritime security and the development of ship autopilot systems. However, existing ship visual datasets primarily focus on ship detection tasks, lacking a fully open-source dataset for multiobject tracking research. Furthermore, current methods often struggle with extracting appearance features under complex sea conditions, varying scales and different ship types, affecting tracking precision. To address these issues, we propose ShipsMOT, a new benchmark dataset containing 121 video sequences with an average of 15.45 s per sequence, covering 15 distinct ship types and a total of 237,999 annotated bounding boxes. Additionally, we propose JDR-CSTrack, a ship multiobject tracking framework that improves feature extraction at different scales by optimising a joint detection and Re-ID network. JDR-CSTrack utilises the fusion of appearance and motion features for multilevel data association, thereby minimising track loss and ID switches. Experimental results confirm that ShipsMOT can serve as a benchmark for future research in ship multiobject tracking and validate the superiority of the proposed JDR-CSTrack framework. The dataset and code can be found on https://github.com/jpj0916/ShipsMOT.
»Cite this article: Luo, F., Jiang, P., Ho, G. T. S., & Zeng, W. (2025). ShipsMOT: A Comprehensive Benchmark and Framework for Multiobject Tracking of Ships. IET Computer Vision, 19(1), e70042.
»DOI:10.1049/cvi2.70042
Authors:
Qianru Zhou, Bai Yin, Baolei Cheng, Yan Wang, Hai Liu, Jianxi Fan.
Abstract:
Large-scale parallel and distributed applications and platforms rely heavily on high-performance network architectures, which greatly emphasize factors including network latency, bandwidth, communication reliability (e.g., connectivity and fault-tolerant routing), etc. Due to the expansion of network scales and the increase of component failures, the demand for the network’s high reliability has grown substantially. Enhancing fault tolerance is the key to improving the network’s reliability, which can be achieved by establishing the connectivity under the “ g -good-neighbor condition”. In the presence of component failures, to maintain normal network operation and minimize communication delays, effective fault-tolerant routing strategies are required. This paper proposes a new class of recursive networks with small diameter, named generalized recursive match networks (GRMNs). GRMNs include well-known networks like generalized hypercubes, BC networks, BCube, HyperX, FBFLY, 2DFB, etc, and other future networks. We investigate the classical connectivity and g -good-neighbor connectivity ( g≥0 ) of the GRMN. In addition, we present an O(N) fault-tolerant unicast routing algorithm with the g -good-neighbor faulty set, called F-TRouting, where N is the number of edges in the GRMN. The algorithm and results can be directly applied to all networks falling within the description of GRMNs, including generalized hypercubes, BC networks, BCube, HyperX, FBFLY, 2DFB, and other future networks. Finally, we give simulation experiment results on the performances of the GRMN and the F-TRouting algorithm, which demonstrate that the GRMN and the F-TRouting have advantageous performances.
»Cite this article: Zhou, Q., Yin, B., Cheng, B., Wang, Y., Liu, H., & Fan, J. (2025). Reliable Communication Performance of Recursive Networks Based on Inter-Subgraph Matching. IEEE Transactions on Networking.
Authors:
Tian, Y., Wu, C.H., Tang, M., & Tian, M.
Abstract:
In this paper, we propose a Bayesian quantile regression (QR) approach to jointly model multivariate ordinal data. Firstly, a multivariate latent variable model is used to link the multivariate ordinal data and latent continuous responses and the multivariate asymmetric Laplace (MAL) distribution is employed to construct the joint QR-based working likelihood for the considered model. Secondly, adaptive- L1/2 penalization priors of regression parameters are incorporated into the working likelihood to implement high-dimensional Bayesian joint QR inference. Markov Chain Monte Carlo (MCMC) algorithm is utilized to derive the fully conditional posterior distributions of all parameters. Thirdly, Bayesian joint relatively QR estimation approach is recommended to result in more efficient estimation results. Finally, Monte Carlo simulation studies and a real instance analysis of multirater agreement data are presented to illustrate the performance of the proposed Bayesian joint relatively QR approach.
»Cite this article: Tian, Y., Wu, C., Tang, M., & Tian, M. (2025). Bayesian joint relatively quantile regression of latent ordinal multivariate linear models with application to multirater agreement analysis. AStA Advances in Statistical Analysis, 109(1), 85-116.
Authors:
Wei, Z., Wu, H., Lin, Z., Wen, Q., Zheng, L., Wen, J., & Liu, H.
Abstract:
The utilization of directional antennas for neighbor discovery in wireless ad hoc networks brings notable benefits, such as extended transmission range, reduced transmission interference, and enhanced antenna gain. However, when nodes use directional antennas for neighbor discovery, the communication range is limited, resulting in a lack of knowledge of potential neighbors. Hence, it is necessary to design a special antenna direction switching strategy for neighbor discovery based on directional antennas. Traditional methods of switching antenna directions are often random or follow predefined sequences, overlooking the historical knowledge of sector exploration for antenna directions. In contrast, existing machine learning approaches aim to leverage observed historical knowledge to adjust antenna directions for faster neighbor discovery. Nonetheless, the latency of neighbor discovery is still high because the node cannot fully utilize the observed historical knowledge (i.e.., only using the knowledge observed by the node in transmission mode, ignoring the knowledge observed by the node in reception mode). Meanwhile, the corresponding reward and penalty mechanisms are still not detailed enough (i.e.., these reward and penalty mechanisms only consider the sectors of discovered and undiscovered neighboring nodes, ignoring the scenario of sectors that have been rewarded). In this paper, the neighbor discovery process is modeled as a reinforcement learning-based learning automaton. We propose an enhanced reinforcement learning-based two-way transmit-receive directional antennas neighbor discovery algorithm, called ERTTND. The algorithm consists of a two-way transmit-receive reinforcement learning mechanism (TTRL) and an enhanced reward-and-penalty mechanism (ERAP). This algorithm leverages insights from nodes in transmission and reception modes to refine their tactical decisions. Then, through an enriched reward-and-penalty framework, nodes optimize their strategies, thus expediting neighbor discovery based on directional antennas in wireless ad hoc networks. Simulation results demonstrate that compared to existing representative algorithms, the proposed ERTTND algorithm can achieve over 30% savings in terms of average discovery delay and energy consumption.
»Cite this article: Wei, Z., Wu, H., Lin, Z., Wen, Q., Zheng, L., Wen, J., & Liu, H. (2025). Enhanced reinforcement learning-based two-way transmit-receive directional antennas neighbor discovery in wireless ad hoc networks. Ad Hoc Networks, 167, 103689.
2024
Paper Title:
Digital transformation for cold chain management in freight forwarding industry
Authors:
Lam, H. Y., & Tang, V.
Abstract:
During the pandemic, the attention and demand for cold chain increased owing to considerable use of low-temperature logistics in transporting perishable goods and vaccines. To ensure the shipping performance for reduced damage, logistics companies are required to track continually and repetitively the status of shipments daily. However, typing various air waybills for searching the shipping status is a cause of frequent errors. Also, tracking the shipping status is labor-intensive, resource intensive, inefficient and repetitive. Moreover, repetitive tasks result in low employee satisfaction. Therefore, robotic process automation (RPA) applications have gained the attention of practitioners in the cold chain logistics industry. This study contributes to (i) determining possible areas requiring automation through the workflow study on cold chain logistics and (ii) streamlining the operation by the develop a robotic process automation bots. A case study tested and evaluated the performance of two unattended RPA bots applied in a freight forwarder company to check shipment status and temperature conditions. The results determined that implementing RPA in the workflow reduces significant data processing time. With the implementation of proposed RPA bots, the company can better comprehend its shipping performance of logistics and can get an immediate notification from RPA bots when an abnormal situation occurs with regard to a shipment.
»Cite this article: Lam, H. Y., & Tang, V. (2023). Digital transformation for cold chain management in freight forwarding industry. International Journal of Engineering Business Management, 15, 18479790231160857.
Paper Title:
Raising logistics performance to new levels through digital transformation
Authors:
Lam, H. Y., Tang, V., & Wong, L.
Abstract:
In view of the higher demand and customer expectations on the speed and accuracy of e-commerce logistics service as well as the repetitive and time-consuming nature of manual order processing operation. This paper proposes a robotic process automation (RPA) model to liberate human resources over time-consuming, inefficient, non-value-added, and repetitive operational processes occupying workforces. This paper proposed an RPA model that integrates three functional RPA bots in (1) tracking order status, (2) capturing order data, and (3) verifying order data to increase the efficiency and accuracy of logistics operation and the order-handling process in small- and medium-sized enterprise (SME) logistics company. A case study was conducted on an SME logistics service providing customers with inbound and outbound operation services. The result demonstrated that the proposed model significantly improves logistics operations performance against a human approach concerning key indicators after implementation. Logistics companies could free up the workforce for value-creating activities in value-added services.
»Cite this article: Lam, H. Y., Tang, V., & Wong, L. (2024). Raising logistics performance to new levels through digital transformation. International Journal of Engineering Business Management, 16, 18479790241231730.
Paper Title:
On-chain and off-chain data management for blockchain-internet of things: a multi-agent deep reinforcement learning approach
Authors:
Tsang, Y. P., Lee, C. K. M., Zhang, K., Wu, C. H., & Ip, W. H.
Abstract:
The emergence of blockchain technology has seen applications increasingly hybridise cloud storage and distributed ledger technology in the Internet of Things (IoT) and cyber-physical systems, complicating data management in decentralised applications (DApps). Because it is inefficient for blockchain technology to handle large amounts of data, effective on-chain and off-chain data management in peer-to-peer networks and cloud storage has drawn considerable attention. Space reservation is a cost-effective approach to managing cloud storage effectively, contrasting with the demand for additional space in real-time. Furthermore, off-chain data replication in the peer-to-peer network can eliminate single points of failure of DApps. However, recent research has rarely discussed optimising on-chain and off-chain data management in the blockchain-enabled IoT (BIoT) environment. In this study, the BIoT environment is modelled, with cloud storage and blockchain orchestrated over the peer-to-peer network. The asynchronous advantage actor-critic algorithm is applied to exploit intelligent agents with the optimal policy for data packing, space reservation, and data replication to achieve an intelligent data management strategy. The experimental analysis reveals that the proposed scheme demonstrates rapid convergence and superior performance in terms of average total reward compared with other typical schemes, resulting in enhanced scalability, security and reliability of blockchain-IoT networks, leading to an intelligent data management strategy.
»Cite this article: Tsang, Y. P., Lee, C. K. M., Zhang, K., Wu, C. H., & Ip, W. H. (2024). On-chain and off-chain data management for blockchain-internet of things: a multi-agent deep reinforcement learning approach. Journal of Grid Computing, 22(1), 16.
Paper Title:
A neighbor discovery protocol with adaptive collision alleviation for wireless robotic networks
Authors:
Yang, J., Tian, Y., & Wu, C. H.
Abstract:
Along with the rapid development of industries and the acceleration of urbanisation, the problem of air pollution is becoming more serious. Exploring the relevant factors affecting air quality and accurately predicting the air quality index are significant in improving the overall environmental quality and realising green economic development. Machine learning algorithms and statistical models have been widely used in air quality prediction and ranking assessment. In this paper, based on daily air quality data for the city of Xi’an, China, from 1 October 2022 to 30 September 2023, we construct support vector regression (SVR), gradient boosting decision tree (GBDT), extreme gradient boosting (XGBoost), random forests (RF), neural network (NN) and long short-term memory (LSTM) models to analyse the influence of the air quality index for Xi’an and to conduct comparative tests. The predicted values and 95% prediction intervals of the AQI for the next 15 days for Xi’an, China, are given based on the Bootstrap-XGBoost algorithm. Further, the ordinal logit regression and ordinal probit regression models are constructed to evaluate and accurately predict the AQI ranks of the data from 1 October 2023 to 15 October 2023 for Xi’an. Finally, this paper proposes some suggestions and policy measures based on the findings of this paper.
»Cite this article: Yang, J., Tian, Y., & Wu, C. H. (2024). Air quality prediction and ranking assessment based on Bootstrap-XGBoost algorithm and ordinal classification models. Atmosphere, 15(8), 925.
Paper Title:
Do Scholars Respond Faster Than Google Trends in Discussing COVID-19 Issues? An Approach to Textual Big Data
Authors:
Lam, B. S. Y., Chu, A. M. Y., Chan, J. N. L., & So, M. K. P.
Abstract:
The COVID-19 pandemic has posed various difficulties for policymakers, such as the identification of health issues, establishment of policy priorities, formulation of regulations, and promotion of economic competitiveness. Evidence-based practices and data-driven decision-making have been recognized as valuable tools for improving the policymaking process. Nevertheless, due to the abundance of data, there is a need to develop sophisticated analytical techniques and tools to efficiently extract and analyze the data. Methods: Using Oxford COVID-19 Government Response Tracker, we categorize the policy responses into 6 different categories: (a) containment and closure, (b) health systems, (c) vaccines, (d) economic, (e) country, and (f) others. We proposed a novel research framework to compare the response times of the scholars and the general public. To achieve this, we analyzed more than 400,000 research abstracts published over the past 2.5 years, along with text information from Google Trends as a proxy for topics of public concern. We introduced an innovative text-mining method: coherent topic clustering to analyze the huge number of abstracts. Results: Our results show that the research abstracts not only discussed almost all of the COVID-19 issues earlier than Google Trends did, but they also provided more in-depth coverage. This should help policymakers identify core COVID-19 issues and act earlier. Besides, our clustering method can better reflect the main messages of the abstracts than a recent advanced deep learning-based topic modeling tool. Conclusion: Scholars generally have a faster response in discussing COVID-19 issues than Google Trends.
»Cite this article: Lam, B. S. Y., Chu, A. M. Y., Chan, J. N. L., & So, M. K. P. (2024). Do Scholars Respond Faster Than Google Trends in Discussing COVID-19 Issues? An Approach to Textual Big Data. Health Data Science, 4, 0116.
»DOI:10.34133/hds.0116
Authors:
Zhang, R., Zhu, D., Wu, C., Xu, J., & Wu, C. H.
Abstract:
Currently, a large number of pornographic images on the Internet severely affect the growth of adolescents. In order to create a healthy and benign online environment, it is necessary to recognize and detect these sensitive images. Current techniques for detecting pornographic content are still in an immature stage, with the key issue being low detection accuracy. To address this problem, this paper proposes a method for detecting pornographic content based on an improved YOLOv8 model. Firstly, InceptionNeXt is introduced into the backbone network to enhance the model’s adaptability to images of different scales and complexities by optimizing feature extraction through parallel branches and deep convolution. Simultaneously, the SPPF module is simplified into the SimCSPSPPF module, which further enhances the effectiveness and diversity of features through improved spatial pyramid pooling and cross-layer feature fusion. Secondly, switchable dilated convolutions are incorporated to improve the adaptability of the C2f enhancement model and enhance the model’s detection capability. Finally, SEAattention is introduced to enhance the model’s ability to capture spatial details. The experiments demonstrate that the model achieves an mAP@0.5 of 79.7% on our self-made sensitive image dataset, which is a significant improvement of 5.9% compared to the previous YOLOv8n network. The proposed method excels in handling complex backgrounds, targets of varying scales, and resource-constrained scenarios, while simultaneously improving the model’s computational efficiency without compromising detection accuracy, making it more advantageous for practical applications.
»Cite this article: Zhang, R., Zhu, D., Wu, C., Xu, J., & Wu, C. H. (2024). Sensitive information detection based on deep learning models. Applied Sciences, 14(17), 7541.
»DOI:10.3390/app14177541
Authors:
Mo, D. Y., Tsang, Y. P., Wang, Y., & Xu, W.
Abstract:
In this study, an online reinforcement learning-based approach and a reinforcement learning with prior knowledge approach are proposed to enhance decision intelligence in inventory management systems for handling nonstationary stochastic market demands in e-commerce environment with crowdsourcing resources. The proposed inventory control policies are designed to solve a multi-period inventory problem with the objectives of optimising inventory-related costs and service levels in the absence of prior information on demand patterns. An experimental analysis reveals that the proposed reinforcement learning-based inventory control policies achieve cost savings and higher service levels across various settings of cost ratios and lead times.
»Cite this article: Mo, D. Y., Tsang, Y. P., Wang, Y., & Xu, W. (2024). Online reinforcement learning-based inventory control for intelligent E-Fulfilment dealing with nonstationary demand. Enterprise Information Systems, 18(2), 2284427.
Authors:
Yang, J., Tian, Y., & Wu, C. H.
Abstract:
Along with the rapid development of industries and the acceleration of urbanisation, the problem of air pollution is becoming more serious. Exploring the relevant factors affecting air quality and accurately predicting the air quality index are significant in improving the overall environmental quality and realising green economic development. Machine learning algorithms and statistical models have been widely used in air quality prediction and ranking assessment. In this paper, based on daily air quality data for the city of Xi’an, China, from 1 October 2022 to 30 September 2023, we construct support vector regression (SVR), gradient boosting decision tree (GBDT), extreme gradient boosting (XGBoost), random forests (RF), neural network (NN) and long short-term memory (LSTM) models to analyse the influence of the air quality index for Xi’an and to conduct comparative tests. The predicted values and 95% prediction intervals of the AQI for the next 15 days for Xi’an, China, are given based on the Bootstrap-XGBoost algorithm. Further, the ordinal logit regression and ordinal probit regression models are constructed to evaluate and accurately predict the AQI ranks of the data from 1 October 2023 to 15 October 2023 for Xi’an. Finally, this paper proposes some suggestions and policy measures based on the findings of this paper.
»Cite this article: Yang, J., Tian, Y., & Wu, C. H. (2024). Air quality prediction and ranking assessment based on Bootstrap-XGBoost algorithm and ordinal classification models. Atmosphere, 15(8), 925.
2023
Paper Title:
Responsive pick face replenishment strategy for stock allocation to fulfil e-commerce order
Authors:
Lam, H. Y., Ho, G. T. S., Mo, D. Y., & Tang, V.
Abstract:
In the rapidly growing e-commerce industry, pallet picking is no longer feasible as the stock keeping units (SKUs) change from the pallet level to the carton or item level. Thus, an effective order-picking process focusing on fast and efficient retrieval of SKUs from shelves is required to fulfil numerous small lot-sized e-commerce orders within a short time. This study investigates a responsive pick face replenishment (RPFR) strategy that divides the high-bay racks in the distribution centres (DCs) into two parts: the upper-deck reserve areas and the pick-face forward areas to improve the operational efficiency in order picking. To address the fluctuating order demand and limited space in the pick-face forward areas, the proposed RPFR system integrates a predictive analytics algorithm with an adaptive network-based fuzzy inference system (ANFIS) and adaptive genetic algorithm-based stock allocation model to generate an optimal stock replenishment plan. By predicting the order demand of each SKU in the next time interval, the types of selected SKUs and their quantities to be loaded into the pick-face forward areas are determined. Numerical experiments are performed to validate the system performance, and comparative analyses are conducted to determine the best parameter settings for the models.
»Cite this article: Lam, H. Y., Ho, G. T. S., Mo, D. Y., & Tang, V. (2023). Responsive pick face replenishment strategy for stock allocation to fulfil e-commerce order. International Journal of Production Economics, 264, 108976.
Paper Title:
Fuzzy superpixel-based image segmentation
Authors:
Ng, T. C., Choy, S. K., Lam, S. Y., & Yu, K. W.
Abstract:
This article presents a multi-phase image segmentation methodology based on fuzzy superpixel decomposition, aggregation and merging. First, a collection of layers of dense fuzzy superpixels is generated by the variational fuzzy decomposition algorithm. Then a layer of refined superpixels is extracted by aggregating various layers of dense fuzzy superpixels using the hierarchical normalized cuts. Finally, the refined superpixels are projected into the low dimensional feature spaces by the multidimensional scaling and the segmentation result is obtained via the mean-shift-based merging approach with the spatial bandwidth adjustment strategy. Our algorithm utilizes the superimposition of fuzzy superpixels to impose more accurate spatial constraints on the final segmentation through the fuzzy superpixel aggregation. The fuzziness of superpixels also provides spatial features to measure affinities between fuzzy superpixels and refined superpixels, and guide the merging process. Comparative experiments with the existing approaches reveal a superior performance of the proposed method.
»Cite this article: Ng, T. C., Choy, S. K., Lam, S. Y., & Yu, K. W. (2023). Fuzzy superpixel-based image segmentation. Pattern Recognition, 134, 109045.
Paper Title:
An automatic speech analytics program for digital assessment of stress burden and psychosocial health
Authors:
Chu, A. M., Lam, B. S., Tsang, J. T., Tiwari, A., Yuk, H., Chan, J. N., & So, M. K.
Abstract:
The stress burden generated from family caregiving makes caregivers particularly prone to developing psychosocial health issues; however, with early diagnosis and intervention, disease progression and long-term disability can be prevented. We developed an automatic speech analytics program (ASAP) for the detection of psychosocial health issues based on clients’ speech. One hundred Cantonese-speaking family caregivers were recruited with the results suggesting that the ASAP can identify family caregivers with low or high stress burden levels with an accuracy rate of 72%. The findings indicate that digital health technology can be used to assist in the psychosocial health assessment. While the conventional method requires rigorous assessments by specialists with multiple rounds of questioning, the ASAP can provide a cost-effective and immediate initial assessment to identify high levels of stress among family caregivers so they can be referred to social workers and healthcare professionals for further assessments and treatments.
»Cite this article: Chu, A. M., Lam, B. S., Tsang, J. T., Tiwari, A., Yuk, H., Chan, J. N., & So, M. K. (2023). An automatic speech analytics program for digital assessment of stress burden and psychosocial health. npj Mental Health Research, 2(1), 15.
Paper Title:
Accelerating L1-penalized expectation maximization algorithm for latent variable selection in multidimensional two-parameter logistic models
Authors:
Shang, L., Xu, P. F., Shan, N., Tang, M. L., & Ho, G. T. S.
Abstract:
One of the main concerns in multidimensional item response theory (MIRT) is to detect the relationship between observed items and latent traits, which is typically addressed by the exploratory analysis and factor rotation techniques. Recently, an EM-based L1-penalized log-likelihood method (EML1) is proposed as a vital alternative to factor rotation. Based on the observed test response data, EML1 can yield a sparse and interpretable estimate of the loading matrix. However, EML1 suffers from high computational burden. In this paper, we consider the coordinate descent algorithm to optimize a new weighted log-likelihood, and consequently propose an improved EML1 (IEML1) which is more than 30 times faster than EML1. The performance of IEML1 is evaluated through simulation studies and an application on a real data set related to the Eysenck Personality Questionnaire is used to demonstrate our methodologies.
»Cite this article: Shang, L., Xu, P. F., Shan, N., Tang, M. L., & Ho, G. T. S. (2023). Accelerating L 1-penalized expectation maximization algorithm for latent variable selection in multidimensional two-parameter logistic models. Plos one, 18(1), e0279918.
Paper Title:
An efficient coprime integer sets based rendezvous algorithm for distributed cognitive radio networks
Authors:
Wei, Z., Liu, Q., Lin, Z., Wen, Q., Wen, J., Liu, H., & Wu, W.
Abstract:
Multichannel blind rendezvous problem is a fundamental problem in cognitive radio networks (CRNs). It refers to how to ensure that any two secondary users can meet on a common available channel at the same time. In this paper, an efficient channel hopping algorithm, called CISR (Coprime Integer Sets based Rendezvous), is designed for asymmetric role model (i.e. users can be divided into two types of roles, say, sender/leader and receiver/follower) in distributed CRNs. CISR can guarantee the rendezvous of the leader and follower and lowers the upper bound of both the maximum time to rendezvous and expected time to rendezvous by carefully constructing two coprime integer sets. The simulation results show that CISR has the near-optimal performance.
»Cite this article: Wei, Z., Liu, Q., Lin, Z., Wen, Q., Wen, J., Liu, H., & Wu, W. (2023). An efficient coprime integer sets based rendezvous algorithm for distributed cognitive radio networks. Electronics Letters, 59(18), e12946.
»DOI:10.1049/ell2.12946
Authors:
Wu, C. H., Lam, H. Y., Kong, A., & Wong, W. L. H.
Abstract:
A majority of government and international organisations have chosen English as their official language since it is the only language in the world with the status of a global language and a world language. One of the key components of learning a language is oral communication; however, students frequently overlook the value of oral communication, which reduces the effectiveness of language acquisition. To enhance the motivation and effectiveness of non-native English speakers to learn and practise English, this study aims to develop an interactive self-learning and self-improvement approach. A well-designed and chatbot-assisted learning environment is designed, using a variety of concepts such as informal learning, language learning, mobile learning, and educational mobile application, and evaluated.
»Cite this article: Wu, C. H., Lam, H. Y., Kong, A., & Wong, W. L. H. (2023). The design and evaluation of a digital learning-based English chatbot as an online self-learning method. International Journal of Engineering Business Management, 15, 18479790231176372.
Authors:
Li, X. J., Tian, G. L., Zhang, M., Ho, G. T. S., & Li, S.
Abstract:
Under-dispersed count data often appear in clinical trials, medical studies, demography, actuarial science, ecology, biology, industry and engineering. Although the generalized Poisson (GP) distribution possesses the twin properties of under- and over-dispersion, in the past 50 years, many authors only treat the GP distribution as an alternative to the negative binomial distribution for modeling over-dispersed count data. To our best knowledge, the issues of calculating maximum likelihood estimates (MLEs) of parameters in GP model without covariates and with covariates for the case of under-dispersion were not solved up to now. In this paper, we first develop a new minimization–maximization (MM) algorithm to calculate the MLEs of parameters in the GP distribution with under-dispersion, and then we develop another new MM algorithm to compute the MLEs of the vector of regression coefficients for the GP mean regression model for the case of under-dispersion. Three hypothesis tests (i.e., the likelihood ratio, Wald and score tests) are provided. Some simulations are conducted. The Bangladesh demographic and health surveys dataset is analyzed to illustrate the proposed methods and comparisons with the existing Conway–Maxwell–Poisson regression model are also presented.
»Cite this article: Li, X. J., Tian, G. L., Zhang, M., Ho, G. T. S., & Li, S. (2023). Modeling under-dispersed count data by the generalized Poisson distribution via two new MM algorithms. Mathematics, 11(6), 1478.
»DOI:10.3390/math11061478
Authors:
Wong, R. S.
Abstract:
Purpose
Despite its significance, research on how attribute framing affects ordering decisions in dual sourcing remains insufficient. Hence, this study investigated the effects of attribute framing in a sourcing task involving certain and uncertain qualities of two suppliers and analysed the role of attention with respect to suppliers’ information in framing effects.
Design/methodology/approach
The impacts of attribute framing on sourcing decisions were demonstrated in two online between-subject (2 × 2 factorial) experimental studies involving professional samples. Study 2 was an eye-tracking experiment.
Findings
In Study 1 (N = 251), participants presented with a “high-quality” rather than a “low-quality” frame made different sourcing decisions, opting for larger percentage of order(s) from a supplier under the “high-quality” frame. This pattern holds true for suppliers who differ in risk. This finding was replicated in Study 2 (N = 129). Attention asymmetry related to the information on supplier quality contributes to this effect. Attention directed towards information regarding the supplier’s quality under a positive frame mediated the relationship between attribute framing and sourcing decisions.
Practical implications
Highlighting the positive attributes of a risky supplier is essential when ordering from the risky supplier is an optimal decision. It is advantageous for suppliers to highlight positive rather than negative attributes when describing the quality of their components against others.
Originality/value
This is the first study to examine the effect of attention on the relationship between attribute framing and dual sourcing. This presents a new behavioural perspective wherein managers’ attention to information plays a vital role.
»Cite this article: Wong, R. S. (2023). An experimental investigation of attribute framing effects on risky sourcing behaviour: the mediating role of attention allocated to suppliers’ quality information. International Journal of Operations & Production Management, 43(13), 205-225.
2022
Paper Title:
Integrated Smart Warehouse and Manufacturing Management with Demand Forecasting in Small-Scale Cyclical Industries
Authors:
Tang, Y. M., Ho, G. T. S., Lau, Y. Y., & Tsui, S. Y.
Abstract:
In the context of the global economic slowdown, demand forecasting, and inventory and production management have long been important topics to the industries. With the support of smart warehouses, big data analytics, and optimization algorithms, enterprises can achieve economies of scale, and balance supply and demand. Smart warehouse and manufacturing management is considered the culmination of recently advanced technologies. It is important to enhance the scalability and extendibility of the industry. Despite many researchers having developed frameworks for smart warehouse and manufacturing management for various fields, most of these models are mainly focused on the logistics of the product and are not generalized to tackle the specific manufacturing problem facing in the cyclical industry. Indeed, the cyclical industry has a key problem: the big risk which high sensitivity poses to the business cycle and economic recession, which is difficult to foresee. Despite many inventory optimization approaches being proposed to optimize the inventory level in the warehouse and facilitate production management, the demand forecasting technique is seldom focused on the cyclic industry. On the other hand, management approaches are usually based on the complex logistics process instead of integrating the inventory level of the stock, which is very crucial to composing smart warehouses and manufacturing. This research study proposed a digital twin framework by integrating the smart warehouse and manufacturing with the roulette genetic algorithm for demand forecasting in the cyclical industry. We also demonstrate how this algorithm is practically implemented for forecasting the demand, sustaining manufacturing optimization, and achieving inventory optimization. We adopted a small-scale textile company case study to demonstrate the proposed digital framework in the warehouse and demonstrate the results of demand forecasting and inventory optimization. Various scenarios were conducted to simulate the results for the digital twin. The proposed digital twin framework and results help manufacturers and logistics companies to improve inventory management. This study has important theoretical and practical significance for the management of the cyclical industry.
»Cite this article: Tang, Y. M., Ho, G. T. S., Lau, Y. Y., & Tsui, S. Y. (2022). Integrated smart warehouse and manufacturing management with demand forecasting in small-scale cyclical industries. Machines, 10(6), 472.
Paper Title:
Design of a Reverse Logistics System with Internet of Things for Service Parts Management
Authors:
Mo, D. Y., Ma, C. Y., Ho, D. C., & Wang, Y.
Abstract:
»Cite this article: Mo, D. Y., Ma, C. Y., Ho, D. C., & Wang, Y. (2022). Design of a reverse logistics system with internet of things for service parts management. Sustainability, 14(19), 12013.
»DOI:10.3390/su141912013
Paper Title:
A forecasting analytics model for assessing forecast error in e-fulfilment performance
Authors:
Ho, G. T. S., Choy, S. K., Tong, P. H., & Tang, V.
Abstract:
Demand forecast methodologies have been studied extensively to improve operations in e-commerce. However, every forecast inevitably contains errors, and this may result in a disproportionate impact on operations, particularly in the dynamic nature of fulfilling orders in e-commerce. This paper aims to quantify the impact that forecast error in order demand has on order picking, the most costly and complex operations in e-order fulfilment, in order to enhance the application of the demand forecast in an e-fulfilment centre.
The paper presents a Gaussian regression based mathematical method that translates the error of forecast accuracy in order demand to the performance fluctuations in e-order fulfilment. In addition, the impact under distinct order picking methodologies, namely order batching and wave picking. As described.
A structured model is developed to evaluate the impact of demand forecast error in order picking performance. The findings in terms of global results and local distribution have important implications for organizational decision-making in both long-term strategic planning and short-term daily workforce planning.
Earlier research examined demand forecasting methodologies in warehouse operations. And order picking and examining the impact of error in demand forecasting on order picking operations has been identified as a research gap. This paper contributes to closing this research gap by presenting a mathematical model that quantifies impact of demand forecast error into fluctuations in order picking performance.
»Cite this article: Ho, G. T. S., Choy, S. K., Tong, P. H., & Tang, V. (2022). A forecasting analytics model for assessing forecast error in e-fulfilment performance. Industrial Management & Data Systems, 122(11), 2583-2608.
Paper Title:
Redeploying excess inventories with lateral and reverse transshipments
Authors:
Mo, D. Y., Wang, Y., Ho, D. C., & Leung, K. H.
Abstract:
Service parts management has the potential to generate high profits for companies that deliver superior service parts services in the after-sale market. However, a big challenge in managing service parts operations is to meet the high expectations of service levels and to reduce excess inventories caused by fluctuating demand and a complex service parts logistics network structure. By expanding the conventional inventory management that passively focuses on the forward and lateral flows of service parts deployment, we propose a crucial but overlooked practice of inventory redeployment as an integral part of the operations that allow the proactive management of lateral and reverse flows of service parts. We formulate the service parts inventory problem with the application of an excess inventory redeployment strategy in a multi-echelon service network as a multi-period integer programming model. This optimisation model is evaluated using a case study of an international company’s service parts operations and demonstrates a higher cost-saving potential. Our novel, integrated approach confers the advantage of redeploying excess inventories in a closed-loop service parts logistics network with a higher cost-saving potential that could not have been achieved in a conventional approach.
»Cite this article: Mo, D. Y., Wang, Y., Ho, D. C., & Leung, K. H. (2022). Redeploying excess inventories with lateral and reverse transshipments. International Journal of Production Research, 60(10), 3031-3046.
Paper Title:
Nonparametric quantile regression with missing data using local estimating equations
Authors:
Wang, C., Tian, M., & Tang, M. L.
Abstract:
In this paper, we propose augmented inverse probability weighted (AIPW) local estimating equations in dealing with missing data in nonparametric quantile regression context. The missing mechanism here is missing at random. To avoid the problem of misspecification, we adopt nonparametric approach to estimate the propensity score and conditional expectations of estimating functions. The asymptotic properties of our proposed estimator are studied. Majorisation–minimisation algorithm is used to circumvent the nonsmoothness of check function at the origin. When it comes to the choice of bandwidth, the theoretical expression of local optimal bandwidth is derived based on asymptotic properties. Moreover, we apply smoothed bootstrap method to obtain the empirical mean square error and use cross-validation to determine the bandwidth in practice. Simulations are conducted to compare the performance of our proposed methods with other existing methods. Finally, we illustrate our methodology with an analysis of non-insulin-dependent diabetes mellitus data set.
»Cite this article: Wang, C., Tian, M., & Tang, M. L. (2022). Nonparametric quantile regression with missing data using local estimating equations. Journal of Nonparametric Statistics, 34(1), 164-186.
2021
Paper Title:
Data analytics and the P2P cloud: an integrated model for strategy formulation based on customer behaviour
Authors:
Lam, H. Y., Tsang, Y. P., Wu, C. H., & Tang, V.
Abstract:
For companies to gain competitive advantage, an effective customer relationship management (CRM) approach is necessary. Based on customer purchase behaviour and ordering patterns, companies can be classified into different categories in terms of providing customised sales and promotions for customers. However, companies that lack an effective CRM strategy can only offer the same sales and marketing strategies to all customers. Furthermore, the traditional approach to managing customers is control via a centralised method, in which the information regarding customer segmentation is not shared among the customer network. Consequently, valuable customers may be neglected, resulting in the loss of customer loyalty and sales orders, and the weakening of trust in the customer–company relationship. This paper designs an integrated data analytic model (IDAM) in a peer-to-peer cloud, integrating RFM-based k-means clustering algorithm, analytical hierarchy processing and fuzzy logic to divide customers into different segments and hence formulate a customised sales strategy. A pilot study of IDAM is conducted in a trading company specialised in providing advanced manufacturing technology to demonstrate how IDAM can be applied to formulate an effective sales strategy to attract customers. Overall, this study explores the effective deployment of CRM into the peer-to-peer cloud so as to facilitate sales strategy formulation and trust between customers and companies in the network.
»Cite this article: Lam, H. Y., Tsang, Y. P., Wu, C. H., & Tang, V. (2021). Data analytics and the P2P cloud: an integrated model for strategy formulation based on customer behaviour. Peer-to-Peer Networking and Applications, 14(5), 2600-2617.
Paper Title:
Exploring the intellectual cores of the blockchain–Internet of Things (BIoT) Available
Authors:
Tsang, Y. P., Wu, C. H., Ip, W. H., & Shiau, W. L.
Abstract:
Due to the rapid growth of blockchain technology in recent years, the fusion of blockchain and the Internet of Things (BIoT) has drawn considerable attention from researchers and industrial practitioners and is regarded as a future trend in technological development. Although several authors have conducted literature reviews on the topic, none have examined the development of the knowledge structure of BIoT, resulting in scattered research and development (R&D) efforts.
This study investigates the intellectual core of BIoT through a co-citation proximity analysis–based systematic review (CPASR) of the correlations between 44 highly influential articles out of 473 relevant research studies. Subsequently, we apply a series of statistical analyses, including exploratory factor analysis (EFA), hierarchical cluster analysis (HCA), k-means clustering (KMC) and multidimensional scaling (MDS) to establish the intellectual core.
»Cite this article: Tsang, Y. P., Wu, C. H., Ip, W. H., & Shiau, W. L. (2021). Exploring the intellectual cores of the blockchain–Internet of Things (BIoT). Journal of Enterprise Information Management, 34(5), 1287-1317.
Paper Title:
Variable selection for ultra-high dimensional quantile regression with missing data and measurement error
Authors:
Bai, Y., Tian, M., Tang, M. L., & Lee, W. Y.
Abstract:
In this paper, we consider variable selection for ultra-high dimensional quantile regression model with missing data and measurement errors in covariates. Specifically, we correct the bias in the loss function caused by measurement error by applying the orthogonal quantile regression approach and remove the bias caused by missing data using the inverse probability weighting. A nonconvex Atan penalized estimation method is proposed for simultaneous variable selection and estimation. With the proper choice of the regularization parameter and under some relaxed conditions, we show that the proposed estimate enjoys the oracle properties. The choice of smoothing parameters is also discussed. The performance of the proposed variable selection procedure is assessed by Monte Carlo simulation studies. We further demonstrate the proposed procedure with a breast cancer data set.
»Cite this article: Bai, Y., Tian, M., Tang, M. L., & Lee, W. Y. (2021). Variable selection for ultra-high dimensional quantile regression with missing data and measurement error. Statistical Methods in Medical Research, 30(1), 129-150.
Paper Title:
Analysis of the Spread of COVID-19 in the USA with a Spatio-Temporal Multivariate Time Series Model
Authors:
Rui, R., Tian, M., Tang, M. L., Ho, G. T. S., & Wu, C. H.
Abstract:
With the rapid spread of the pandemic due to the coronavirus disease 2019 (COVID-19), the virus has already led to considerable mortality and morbidity worldwide, as well as having a severe impact on economic development. In this article, we analyze the state-level correlation between COVID-19 risk and weather/climate factors in the USA. For this purpose, we consider a spatio-temporal multivariate time series model under a hierarchical framework, which is especially suitable for envisioning the virus transmission tendency across a geographic area over time. Briefly, our model decomposes the COVID-19 risk into: (i) an autoregressive component that describes the within-state COVID-19 risk effect; (ii) a spatiotemporal component that describes the across-state COVID-19 risk effect; (iii) an exogenous component that includes other factors (e.g., weather/climate) that could envision future epidemic development risk; and (iv) an endemic component that captures the function of time and other predictors mainly for individual states. Our results indicate that maximum temperature, minimum temperature, humidity, the percentage of cloud coverage, and the columnar density of total atmospheric ozone have a strong association with the COVID-19 pandemic in many states. In particular, the maximum temperature, minimum temperature, and the columnar density of total atmospheric ozone demonstrate statistically significant associations with the tendency of COVID-19 spreading in almost all states. Furthermore, our results from transmission tendency analysis suggest that the community-level transmission has been relatively mitigated in the USA, and the daily confirmed cases within a state are predominated by the earlier daily confirmed cases within that state compared to other factors, which implies that states such as Texas, California, and Florida with a large number of confirmed cases still need strategies like stay-at-home orders to prevent another outbreak.
»Cite this article: Rui, R., Tian, M., Tang, M. L., Ho, G. T. S., & Wu, C. H. (2021). Analysis of the spread of COVID-19 in the USA with a spatio-temporal multivariate time series model. International Journal of Environmental Research and Public Health, 18(2), 774.
Paper Title:
Immersive Learning Design for Technology Education: A Soft Systems Methodology
Authors:
Wu, C. H., Tang, Y. M., Tsang, Y. P., & Chau, K. Y.
Abstract:
Science, technology, engineering and mathematics (STEM) education is a globalized trend of equipping students to facilitate technological and scientific developments. Among STEM education, technology education (TE) plays a significant role in teaching applied knowledge and skills to create and add value to systems and products. In higher education, the learning effectiveness of the TE assisted by the immersive technologies is an active research area to enhance the teaching quality and learning performance. In this study, a taught subject of radio frequency identification (RFID) assisted by using mixed reality technologies in a higher education institution was examined, while the soft systems methodology (SSM) was incorporated to evaluate the changes in learning performance. Under the framework of SSM, stakeholders’ perceptions toward immersive learning and RFID education are structured. Thus, a rich picture for teaching activities is established for subject control, monitoring, and evaluation. Subsequently, the design of TE does not only satisfy the students’ needs but also requirements from teachers, industries, and market trends. Finally, it is found that SSM is an effective approach in designing courses regarding hands-on technologies, and the use of immersive technologies improves the learning performance for acquiring fundamental knowledge and application know-how.
»Cite this article: Wu, C. H., Tang, Y. M., Tsang, Y. P., & Chau, K. Y. (2021). Immersive learning design for technology education: A soft systems methodology. Frontiers in psychology, 12, 745295.
Paper Title:
An Alternative Explanation for Attribute Framing and Spillover Effects in Multidimensional Supplier Evaluation and Supplier Termination: Focusing on Asymmetries in Attention
Authors:
Wong, R. S.
Abstract:
Prior research on attribute framing has focused on a single-dimensional evaluation. However, supplier performance is usually evaluated against multiple dimensions. This research examines framing effects on multidimensional supplier evaluation and investigates the interplay between framing and attention. Three online experiments involving professional participants were conducted in which a supplier performance dimension was framed either positively or negatively. The results from Studies 1A (N = 113) and 1B (N = 82) demonstrated that framing a performance dimension of supplier negatively (versus positively) lowered evaluation of other nonframed dimensions even when these dimensions were presented identically. Framing a dimension negatively also impacted selection decisions, making participants more likely to discontinue purchasing from the supplier. These findings suggest that supplier evaluation and selection are susceptible to framing effects. Study 2 used a web-deployed eye-tracking experiment (N = 62) to elucidate the role of attention in framing effects. I found that performance information, when framed negatively rather than positively, received more attention. Also, attention to the framed performance dimension partially mediated the relationship between attribute framing and the evaluation of the framed dimension. This finding provides a fuller understanding of the cause of framing effect. An important managerial implication is that even when a performance indicator is a quantitative, tangible attribute, it is prudent for supply managers to reframe this indicator and check whether the same evaluation and selection decisions are made. Other managerial implications and directions for future research are also discussed.
»Cite this article: Wong, R. S. (2021). An alternative explanation for attribute framing and spillover effects in multidimensional supplier evaluation and supplier termination: Focusing on asymmetries in attention. Decision Sciences, 52(1), 262-282.
»DOI: 10.1111/deci.12435
Paper Title:
Combined soft measurement on key indicator parameters of new competitive advantages for China’s export
Authors:
Wang, T., Zuo, H., Wu, C. H., & Hu, B.
Abstract:
The estimation of the difference between the new competitive advantages of China’s export and the world’s trading powers have been the key measurement problems in China-related studies. In this work, a comprehensive evaluation index system for new export competitive advantages is developed, a soft-sensing model for China’s new export competitive advantages based on the fuzzy entropy weight analytic hierarchy process is established, and the soft-sensing values of key indexes are derived. The obtained evaluation values of the main measurement index are used as the input variable of the fuzzy least squares support vector machine, and a soft-sensing model of the key index parameters of the new export competitive advantages of China based on the combined soft-sensing model of the fuzzy least squares support vector machine is established. The soft-sensing results of the new export competitive advantage index of China show that the soft measurement model developed herein is of high precision compared with other models, and the technical and brand competitiveness indicators of export products have more significant contributions to the new competitive advantages of China’s export, while the service competitiveness indicator of export products has the least contribution to new competitive advantages of China’s export.
»Cite this article: Wang, T., Zuo, H., Wu, C. H., & Hu, B. (2021). Combined soft measurement on key indicator parameters of new competitive advantages for China’s export. Financial Innovation, 7(1), 50.
Paper Title:
Cervical cell classification based on the CART feature selection algorithm
Authors:
Dong, N., Zhai, M. D., Zhao, L., & Wu, C. H.
Abstract:
In recent years, conventional artificial method leads to low efficiency in the classification of cervical cell, which requires professional completion. Therefore, the classification process is increasingly dependent on artificial intelligence. The traditional image classification method needs to extract a large number of features. Redundant features cause the recognition speed to be slow, and influence the recognition effect. To address these problems and obtain the highest recognition accuracy with the least number of features, this paper proposes a machine learning method based on feature selection algorithm for cervical cell classification. Firstly, we introduced classification and regression trees (CART) for cell feature selection, which reduces the dimension of input feature attributes. Subsequently, particle swarm optimization (PSO) was used to optimize the hyperparameters of support vector machine (SVM) in this paper, making the SVM model better for classification. Finally, the Herlev dataset was introduced to verify the classification performance. The experimental results show that the proposed algorithm can extract accurate and effective features and obtain high classification accuracy, thus verifying the effectiveness of the proposed algorithm. Moreover, the network structure of the proposed algorithm is relatively simple with a low computation cost, which makes it feasible of further extension to the classification application of other cancer cells.
»Cite this article: Dong, N., Zhai, M. D., Zhao, L., & Wu, C. H. (2021). Cervical cell classification based on the CART feature selection algorithm. Journal of Ambient Intelligence and Humanized Computing, 12(2), 1837-1849.
Paper Title:
Fuzzy Bit-Plane-Dependence Region Competition
Authors:
Choy, S., Ng, T., Yu, C., & Lam, B.
Abstract:
This paper presents a novel variational model based on fuzzy region competition and statistical image variation modeling for image segmentation. In the energy functional of the proposed model, each region is characterized by the pixel-level color feature and region-level spatial/frequency information extracted from various image domains, which are modeled by the windowed bit-plane-dependence probability models. To efficiently minimize the energy functional, we apply an alternating minimization procedure with the use of Chambolle’s fast duality projection algorithm, where the closed-form solutions of the energy functional are obtained. Our method gives soft segmentation result via the fuzzy membership function, and moreover, the use of multi-domain statistical region characterization provides additional information that can enhance the segmentation accuracy. Experimental results indicate that the proposed method has a superior performance and outperforms the current state-of-the-art superpixel-based and deep-learning-based approaches.
»Cite this article: Choy, S., Ng, T., Yu, C., & Lam, B. (2021). Fuzzy Bit-Plane-Dependence Region Competition. Mathematics, 9(19), 2392.
»DOI: 10.3390/math9192392
Paper Title:
Kernel density-based likelihood ratio tests for linear regression models
Authors:
Yan, F., Xu, Q. S., Tang, M. L., & Chen, Z.
Abstract:
In this article, we develop a so-called profile likelihood ratio test (PLRT) based on the estimated error density for the multiple linear regression model. Unlike the existing likelihood ratio test (LRT), our proposed PLRT does not require any specification on the error distribution. The asymptotic properties are developed and the Wilks phenomenon is studied. Simulation studies are conducted to examine the performance of the PLRT. It is observed that our proposed PLRT generally outperforms the existing LRT, empirical likelihood ratio test and the weighted profile likelihood ratio test in sense that (i) its type I error rates are closer to the prespecified nominal level; (ii) it generally has higher powers; (iii) it performs satisfactorily when moments of the error do not exist (eg, Cauchy distribution); and (iv) it has higher probability of correctly selecting the correct model in the multiple testing problem. A mammalian eye gene expression dataset and a concrete compressive strength dataset are analyzed to illustrate our methodologies.
»Cite this article: Yan, F., Xu, Q. S., Tang, M. L., & Chen, Z. (2021). Kernel density‐based likelihood ratio tests for linear regression models. Statistics in Medicine, 40(1), 119-132.
»DOI: 10.1002/sim.8765
2020
Paper Title:
A Fuzzy-Based Product Life Cycle Prediction for Sustainable Development in the Electric Vehicle Industry
Authors:
Tsang, Y. P., Wong, W. C., Huang, G. Q., Wu, C. H., Kuo, Y. H., & Choy, K. L.
Abstract:
The development of electric vehicles (EVs) has drawn considerable attention to the establishment of sustainable transport systems to enable improvements in energy optimization and air quality. EVs are now widely used by the public as one of the sustainable transportation measures. Nevertheless, battery charging for EVs create several challenges, for example, lack of charging facilities in urban areas and expensive battery maintenance. Among various components in EVs, the battery pack is one of the core consumables, which requires regular inspection and repair in terms of battery life cycle and stability. The charging efficiency is limited to the power provided by the facilities, and therefore the current business model for EVs is not sustainable. To further improve its sustainability, plug-in electric vehicle battery pack standardization (PEVBPS) is suggested to provide a uniform, standardized and mobile EV battery that is managed by centralized service providers for repair and maintenance tasks. In this paper, a fuzzy-based battery life-cycle prediction framework (FBLPF) is proposed to effectively manage the PEVBPS in the market, which integrates the multi-responses Taguchi method (MRTM) and the adaptive neuro-fuzzy inference system (ANFIS) as a whole for the decision-making process. MRTM is formulated based on selection of the most relevant and critical input variables from domain experts and professionals, while ANFIS takes part in time-series forecasting of the customized product life-cycle for demand and electricity consumption. With the aid of the FPLCPF, the revolution of the EV industry can be revolutionarily boosted towards total sustainable development, resulting in pro-active energy policies in the PEVBPS eco-system.
»Cite this article: Tsang, Y. P., Wong, W. C., Huang, G. Q., Wu, C. H., Kuo, Y. H., & Choy, K. L. (2020). A fuzzy-based product life cycle prediction for sustainable development in the electric vehicle industry. Energies, 13(15), 3918.
»DOI:10.3390/en13153918
Paper Title:
Variable Screening for Near Infrared (NIR) Spectroscopy Data Based on Ridge Partial Least Squares Regression
Authors:
Zhao, N., Xu, Q., Tang, M. L., & Wang, H.
Abstract:
Aim and Objective: Near Infrared (NIR) spectroscopy data are featured by few dozen to many thousands of samples and highly correlated variables. Quantitative analysis of such data usually requires a combination of analytical methods with variable selection or screening methods. Commonly-used variable screening methods fail to recover the true model when (i) some of the variables are highly correlated, and (ii) the sample size is less than the number of relevant variables. In these cases, Partial Least Squares (PLS) regression based approaches can be useful alternatives. Materials and Methods: In this research, a fast variable screening strategy, namely the preconditioned screening for ridge partial least squares regression (PSRPLS), is proposed for modelling NIR spectroscopy data with high-dimensional and highly correlated covariates. Under rather mild assumptions, we prove that using Puffer transformation, the proposed approach successfully transforms the problem of variable screening with highly correlated predictor variables to that of weakly correlated covariates with less extra computational effort. Results: We show that our proposed method leads to theoretically consistent model selection results. Four simulation studies and two real examples are then analyzed to illustrate the effectiveness of the proposed approach. Conclusion: By introducing Puffer transformation, high correlation problem can be mitigated using the PSRPLS procedure we construct. By employing RPLS regression to our approach, it can be made more simple and computational efficient to cope with the situation where model size is larger than the sample size while maintaining a high precision prediction.
»Cite this article: Zhao, N., Xu, Q., Tang, M. L., & Wang, H. (2020). Variable screening for near infrared (NIR) spectroscopy data based on ridge partial least squares regression. Combinatorial Chemistry & High Throughput Screening, 23(8), 740-756.
Paper Title:
A Fast Binary Quadratic Programming Solver Based on Stochastic Neighborhood Search
Authors:
Lam, B. S. Y., & Liew, A. W. C.
Abstract:
Many image processing and pattern recognition problems can be formulated as binary quadratic programming (BQP) problems. However, solving a large BQP problem with a good quality solution and low computational time is still a challenging unsolved problem. Current methodologies either adopt an independent random search in a semi-definite space or perform search in a relaxed biconvex space. However, the independent search has great computation cost as many different trials are needed to get a good solution. The biconvex search only searches the solution in a local convex ball, which can be a local optimal solution. In this paper, we propose a BQP solver that alternatingly applies a deterministic search and a stochastic neighborhood search. The deterministic search iteratively improves the solution quality until it satisfies the KKT optimality conditions. The stochastic search performs bootstrapping sampling to the objective function constructed from the potential solution to find a stochastic neighborhood vector. These two steps are repeated until the obtained solution is better than many of its stochastic neighborhood vectors. We compare the proposed solver with several state-of-the-art methods for a range of image processing and pattern recognition problems. Experimental results showed that the proposed solver not only outperformed them in solution quality but also with the lowest computational complexity.
»Cite this article: Lam, B. S. Y., & Liew, A. W. C. (2020). A fast binary quadratic programming solver based on stochastic neighborhood search. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(1), 32-49.
Paper Title:
Variational Fuzzy Superpixel Segmentation
Authors:
Ng, T. C., & Choy, S. K.
Abstract:
This article presents a novel variational model based on fuzzy clustering and total variation regularization for superpixel segmentation. Compared with the classical hard-labeled methodologies, our approach gives soft results via the fuzzy membership function, and moreover, the use of total variation provides additional information that can enhance the superpixel regularity, which in turn improves the segmentation performance. To efficiently minimize the energy functional of the proposed model, we adopt an alternating direction method of multipliers with the modified Chambolle’s fast duality projection algorithm. Our algorithm can generate regular and compact superpixels with high segmentation accuracy, satisfactory boundary adherence, and low computational cost. Comparative experimental results with the current state-of-the-art approaches reveal the superior performance of the proposed method.
»Cite this article: Ng, T. C., & Choy, S. K. (2020). Variational fuzzy superpixel segmentation. IEEE Transactions on Fuzzy Systems, 30(1), 14-26.
Paper Title:
Inception v3 based cervical cell classification combined with artificially extracted features
Authors:
Dong, N., Zhao, L., Wu, C. H., & Chang, J. F.
Abstract:
Traditional cell classification methods generally extract multiple features of the cell manually. Moreover, the simple use of artificial feature extraction methods has low universality. For example, it is unsuitable for cervical cell recognition because of the complexity of the cervical cell texture and the large individual differences between cells. Using the convolutional neural network classification method is a good way to solve this problem. However, although the cell features can be extracted automatically, the cervical cell domain knowledge will be lost, and the corresponding features of different cell types will be missing; hence, the classification effect is not sufficiently accurate. Aiming at addressing the limitations of the two mentioned classification methods, this paper proposes a cell classification algorithm that combines Inception v3 and artificial features, which effectively improves the accuracy of cervical cell recognition. In addition, to address the under-fitting problem and carry out effective deep learning training with a relatively small amount of medical data, this paper inherits the strong learning ability from transfer learning, and achieves accurate and effective cervical cell image classification based on the Herlev dataset. Using this method, an accuracy of more than 98% is achieved, providing an effective framework for computer aided diagnosis of cervical cancer. The proposed algorithm has good universality, low complexity, and high accuracy, rendering it suitable for further extension and application to the classification of other types of cancer cells.
»Cite this article: Dong, N., Zhao, L., Wu, C. H., & Chang, J. F. (2020). Inception v3 based cervical cell classification combined with artificially extracted features. Applied Soft Computing, 93, 106311.
Refereed Conference Papers
2026
Paper Title:
Indoor GIS Applications in Smart Buildings: A Case Study and User-Centric Analysis
Authors:
Peng, J. X., Mo, D. Y., Wu, C. H., & Chan, C. Y.
Abstract:
The rapid advancement of digital technologies and the development of smart cities have accelerated digital transformation across industries, with smart buildings serving as a critical foundation for this transition. Indoor Geographic Information System (GIS) technologies play a pivotal role in smart buildings, offering advanced spatial analysis capabilities to enhance facility management, improve energy efficiency, and provide personalised services such as indoor navigation and real-time positioning. Despite their potential, the application of indoor GIS in smart building development remains underexplored. This study investigates the potential of indoor GIS applications in smart buildings, using a local university campus as a case study. By integrating Axiomatic Design (AD) theory and empirical models, this research evaluates the preferences and expectations of campus stakeholders regarding indoor GIS features. Regression analysis of the results highlights significant demand for applications such as personal location sharing, real-time crowd monitoring, and indoor navigation. The findings provide actionable insights into user-centric design for indoor GIS and its extended applications, which contribute to integrating digital technologies into a smart building ecosystem. This study offers a practical framework for advancing smart building capabilities, aligning with broader sustainability and digital transformation goals.
»Cite this article: Peng, J.X., Mo, D.Y., Wu, C.H., Chan, C.Y. (2026). Indoor GIS Applications in Smart Buildings: A Case Study and User-Centric Analysis. In: Lee, C.K.M., Wu, J.C., Tsang, P.Y., Keung, D.K. (eds) Smart Production for Sustainability. ICPR-APR 2025. Responsible Innovation in Industry. Springer, Singapore. https://doi.org/10.1007/978-981-95-5819-3_2
2025
Paper Title:
Enhancing Personalized Shopping in the Metaverse: A Fuzzy Multi-criteria Decision Model for Luxury Retail Layout Optimization
Authors:
Lam, H. Y., Ho, G. T. S., Wong, L., & Mo, D. Y.
Abstract:
Globalization and market competition has transformed luxury retail from a traditional to an ecommerce business model. However, providing customers with exclusive personalization and immersive experience through an e-commerce platform is a challenge. To provide a better shopping experience, luxury retailers have been developing virtual stores using the metaverse. Using real-time reconfigured product layouts, the retailer can dynamically change the display of a large number of products for customers. However, luxury retailers have difficulty changing the evaluation criteria for personalized and data-driven customer experiences because of the changing environment from two- to three-dimensional. Therefore, this study proposes a fuzzy-based metaverse retail layout optimization model to enhance dynamic product layouts in luxury retail environments by considering six real-time behavioral criteria. The results show that the proposed model can guide strategic product placement and improve the virtual shopping experience of customers. This study enhances metaverse commerce, luxury retail, and intelligent decision systems that offer a practical approach to digital transformation.
»Cite this article: Lam, H. Y., Ho, G. T. S., Wong, L., & Mo, D. Y. (2025, December). Enhancing Personalized Shopping in the Metaverse: A Fuzzy Multi-criteria Decision Model for Luxury Retail Layout Optimization. In 2025 IEEE International Conference on Industrial Engineering and Engineering Management (IEEM) (pp. 1326-1330). IEEE.
Paper Title:
Deep Reinforcement Learning for Order-picking Optimization with Dynamic Order Arrivals in Express Centers
Authors:
Mo, D. Y., Tsang, Y. P., Wang, Y., Chung, K. T., & Lee, C. K. M.
Abstract:
Quick commerce relies on ultra-fast and low-cost order fulfilment, yet its primary bottleneck is order picking at express logistics centers, where fragmented and time-critical requests must be consolidated and dispatched within hours. To address this challenge, this study proposes a deep reinforcement learning (DRL) approach that learns optimal postponement policies for express logistics center operations with dynamic order arrivals. Through simulation experiments, the proposed DRL approach is shown to achieve a cost saving of 14.6% compared to other heuristics under standard capacity conditions, and the cost-saving potential is inversely proportional to the available capacity. These results provide practical guidelines for optimizing the order-picking process using DRL technology.
»Cite this article: Mo, D. Y., Tsang, Y. P., Wang, Y., Chung, K. T., & Lee, C. K. M. (2025, December). Deep Reinforcement Learning for Order-picking Optimization with Dynamic Order Arrivals in Express Centers. In 2025 IEEE International Conference on Industrial Engineering and Engineering Management (IEEM) (pp. 0068-0072). IEEE.
Paper Title:
Two-Sided Matching for Batch-Aware LLM Request Scheduling in Edge Networks
Authors:
Li, T., & Gong, Y.
Abstract:
Edge Large Language Model (LLM) inference offers reduced latency and enhanced privacy compared to cloud-based approaches. Current inference engines utilize batching mechanisms for computational efficiency. However, static batching creates prompt interdependencies, leading to inefficient memory usage and prolonged processing times due to batch heterogeneity. Existing request scheduling solutions optimize single-server batching but cannot coordinate across multiple edge servers or handle resource constraints at edge. In this paper, we propose a batch-aware request scheduling framework formulated as a two-sided matching game, where batch composition affects individual inference performance through peer effects. We design novel utility functions for users and edge servers based on inference quality, memory and computing capability, construct preference lists, and establish stable assignments via Gale-Shapley algorithm. Based on this, beneficial swaps further optimize batch homogeneity while preserving service quality. Extensive simulations demonstrate substantial improvements in service quality, consistently outperforming baselines across varying request volumes and resource constraints.
»Cite this article: Li, T., & Gong, Y. (2025, October). Two-Sided Matching for Batch-Aware LLM Request Scheduling in Edge Networks. In 2025 IEEE 50th Conference on Local Computer Networks (LCN) (pp. 1-7). IEEE.
Paper Title:
Can Kolmogorov-Arnold Network (KAN) Replace Multi-layer Perception (MLP) in Reinforcement Learning for Stochastic Inventory Control?
Authors:
Tsang, Y. P., Mo, D. Y., Chung, K. T., & Lee, C. K. M.
Abstract:
Effective inventory control is pivotal in supply chain management for maintaining desired service levels while minimizing costs. However, real-world business scenarios often present challenges such as uncertain demand and lead times, making inventory management particularly complex. In recent years, reinforcement learning (RL) has emerged as a promising approach for developing adaptive and efficient inventory control policies in these uncertain environments. Despite the demonstrated performance of RL-based solutions, their limited interpretability poses challenges for supply chain managers, who may find it difficult to trust in replicating the promising results in future applications. To address the interpretability concerns, the Kolmogorov-Arnold Network (KAN) has been proposed as an alternative to the traditionally employed Multi-Layer Perceptron (MLP) within RL frameworks. However, the comparative performance of KAN versus MLP in RL deployments for inventory control remains unexplored. This study is motivated to bridge this knowledge gap by systematically comparing MLP-based and KAN-based RL deployments in a stochastic inventory control problem. Through comprehensive computational experiments, we demonstrate that MLP-based approaches, specifically the Double Deep Q-Network (DDQN), generally outperform KAN-based models in terms of overall inventory management efficiency. Notably, as lead time increases, the performance disparity between MLP and KAN diminishes, with their 95% confidence intervals overlapping, indicating comparable performance under longer lead times. These findings suggest that while MLPs may offer superior performance in standard settings, KANs hold potential in scenarios with extended lead times. Consequently, this research contributes to the field by providing empirical evidence to inform the selection of neural network architectures in RL-based inventory control systems. It supports the ongoing development of more interpretable and effective RL variants tailored to address the complexities and uncertainties inherent in real-world inventory management.
»Cite this article: Tsang, Y. P., Mo, D. Y., Chung, K. T., & Lee, C. K. M. (2025, April). Can Kolmogorov-Arnold Network (KAN) Replace Multi-layer Perception (MLP) in Reinforcement Learning for Stochastic Inventory Control?. In Proceedings of the 2025 9th International Conference on Intelligent Systems, Metaheuristics & Swarm Intelligence (pp. 58-64).
»DOI: 10.1145/3760622.3760623
Paper Title:
Utilizing Quantum Annealing to Address Vehicle Routing Challenges in Cold Chain Logistics
Authors:
Chow, E. W. H., Ho, G. T. S., Tang, V., & Tam, M. M. F.
Abstract:
In the current competitive industrial environment, the study of NP-hard issues is becoming more and more significant. In logistics, the Vehicle Routing Problem (VRP) is one of the popular challenges, especially in cold chain logistics, which requires the management of perishable goods. This problem involves identifying the optimal paths for vehicles responsible for transporting goods to different locations. Due to its combinatorial nature and the complexity of finding optimal solutions, it is identified as an NP-hard problem that requires an unreasonable time frame to solve. Recently, quantum computing has proven to be a promising approach for addressing some of the complex challenges. Among quantum methodologies, quantum annealing (QA) has demonstrated the potential to accelerate computation times in solving optimization problems. By leveraging the principles of quantum tunneling and superposition, QA can explore solution spaces more efficiently. Therefore, this study explores the application of QA to address VRP in cold chain logistics, aiming to develop solutions that minimize travel distances during deliveries. The case study demonstrates that QA can achieve, on average, at least 85% of optimal performance while providing approximately 95% faster run times compared to classical genetic algorithms (GA). This contributes to greater efficiency in cold chain logistics operations.
»Cite this article: Chow, E. W. H., Ho, G. T. S., Tang, V., & Tam, M. M. F. (2025, April). Utilizing Quantum Annealing to Address Vehicle Routing Challenges in Cold Chain Logistics. In Proceedings of the 2025 9th International Conference on Intelligent Systems, Metaheuristics & Swarm Intelligence (pp. 137-142).
»DOI: 10.1145/3760622.376062
2024
Paper Title:
Metaverse-enabled Responsive Data Analytics Model for Enhancing Customers’ Online Experience in Luxury Retail
Authors:
Lam, H. Y., & Wong, L.
Abstract:
Luxury retail sales have declined because of restrictions and consumer concerns during the COVID-19 pandemic. To adapt to these challenges, luxury retailers have transformed their businesses from offline platforms to ecommerce platforms. However, purchasing luxury goods online presents significant challenges for customers because they cannot evaluate products through sensory touch and tangible experiences. To address this issue, luxury retailers are turning to the metaverse. Luxury retailers can create immersive and interactive online experiences by developing virtual shopping platforms in metaverse environments. This integration of online and brick-and-mortar channels provides personalized customer interactions, facilitating virtual product trials, inquiries, and recommendations. The aim of this study was to design a metaverse-enabled responsive data analytics model to collect and analyze large amounts of data to uncover valuable insights into customer behavior and purchasing patterns, aiding decision-making in customer relationship management. A case study was conducted to verify the feasibility of the proposed model. The results show that customer behavior is related to the consumption level, service satisfaction, and eye fixation time.
»Cite this article: Lam, H. Y., & Wong, L. (2024, December). Metaverse-enabled Responsive Data Analytics Model for Enhancing Customers’ Online Experience in Luxury Retail. In 2024 IEEE International Conference on Industrial Engineering and Engineering Management (IEEM) (pp. 1402-1406). IEEE.
Paper Title:
An Intelligent Decision Support Model for Sustainable Automated E-fulfillment Centers Through Fuzzy Association Rules Mining
Authors:
Ho, G. T.S., & Tang, V.
Abstract:
Recently, the logistics sector has come to recognize the importance of sustainability, especially with regard to automated e-fulfillment centers that mainly rely on automation and robotics technology. Practitioners are increasingly recognizing the importance of efficiently maintaining order fulfillment while minimizing the costs associated with renewable energy. Thus, practitioners need decisional support to optimize their renewable energy resources and prevent wastage, all while ensuring effective operations. This study uses fuzzy association rules mining (FARM) to reveal links that are hidden between environmental data and the production of renewable energy. “IF-then” rules will be produced by the FARM, providing decision support in the field. The proposed model focuses on solar power as a case study. By utilizing the extracted rules, users can identify the most influential factors that impact solar power generation. By targeting those crucial factors, practitioners can effectively determine the ideal quantity of solar panels, allowing them to prevent the accumulation of unused solar panels while still maintaining optimal operational performance.
»Cite this article: Ho, G. T., & Tang, V. (2024, December). An Intelligent Decision Support Model for Sustainable Automated E-fulfillment Centers Through Fuzzy Association Rules Mining. In 2024 IEEE International Conference on Industrial Engineering and Engineering Management (IEEM) (pp. 1397-1401). IEEE.
Paper Title:
Age of Collection with Network-Coded Multiple Access: An Experimental Study
Paper Title:
Utilizing Quantum Annealing to Address Vehicle Routing Challenges in Cold Chain Logistics
Authors:
Lai, Y., Liu, H., Chan, T. T., Pan, H., & Jiang, C.
Abstract:
This paper studies information freshness in collaborative surveillance scenarios operated with non-orthogonal multiple access (NOMA), where each monitoring device observes a portion of a common target and reports its latest status on the target to a common access point (AP) to recover the complete observation. We use age of collection (AoC) as a metric of information freshness. Unlike the conventional age of information (AoI) metric, the instantaneous AoC decreases only when the AP receives all partial updates from different devices (i.e., successfully receives a “joint” update). Conventional NOMA schemes typically use multiuser decoding (MUD) techniques to decode update messages from different devices. However, MUD does not work well when the signal-to-noise ratios (SNRs) of different NOMA users are (nearly) balanced. Therefore, we consider network-coded multiple access (NCMA), an advanced NOMA scheme that integrates MUD with physical-layer network coding (PNC). PNC is a technique that turns wireless interferences into useful network-coded information, which works well even when the SNRs of different users do not differ much. Experimental results on software-defined radios indicate that NCMA is a practical solution for achieving low average AoC under different channel conditions. This is the first study to show that NCMA, thanks to the combination of MUD and PNC, can receive joint update messages in a shorter period of time, thus significantly reducing the average AoC of the system.
»Cite this article: Lai, Y., Liu, H., Chan, T. T., Pan, H., & Jiang, C. (2024, June). Age of Collection with Network-Coded Multiple Access: An Experimental Study. In 2024 IEEE 99th Vehicular Technology Conference (VTC2024-Spring) (pp. 1-6). IEEE.
Paper Title:
Personalized Federated Deep Reinforcement Learning for Heterogeneous Edge Content Caching Networks
Authors:
Li, Z., Li, T., Liu, H., & Chan, T. T.
Abstract:
Proactive caching is essential for minimizing latency and improving Quality of Experience (QoE) in multi-server edge networks. Federated Deep Reinforcement Learning (FDRL) is a promising approach for developing cache policies tailored to dynamic content requests. However, FDRL faces challenges such as an expanding caching action space due to increased content numbers and difficulty in adapting global information to heteroge-neous edge environments. In this paper, we propose a Personalized Federated Deep Reinforcement Learning framework for Caching, called PF-DRL-Ca, with the aim to maximize system utility while satisfying caching capability constraints. To manage the expanding action space, we employ a new DRL algorithm, Multi-head Deep Q-Network (MH-DQN), which reshapes the action output layers of DQN into a multi-head structure where each head generates a sub-dimensional action. We next integrate the proposed MH-DQN into a personalized federated training framework, employing a layer-wise approach for training to derive a personalized model that can adapt to heterogeneous environments while exploiting the global information to accelerate learning convergence. Our extensive experimental results demonstrate the superiority of MH-DQN over traditional DRL algorithms on a single server, as well as the advantages of the personal federated training architecture compared to other frameworks.
»Cite this article: Li, Z., Li, T., Liu, H., & Chan, T. T. (2024, October). Personalized federated deep reinforcement learning for heterogeneous edge content caching networks. In 2024 22nd International Symposium on Modeling and Optimization in Mobile, Ad Hoc, and Wireless Networks (WiOpt) (pp. 313-320). IEEE.
»Electronic ISBN:978-3-903176-65-2
Paper Title:
Peak Age of Collection in Coordinated Direct and Relay Transmission with Physical-Layer Network Coding
Authors:
Meng, G., Liu, H., & Chan, T. T.
Abstract:
This paper investigates the information freshness of joint status updates, quantified by the age of collection (AoC), in uplink coordinated direct and relay transmission (CDRT). In an uplink CDRT setup, a direct sensor communicates directly with the destination, while a relay-aided sensor relies on a decode-and-forward relay. The update packet of each sensor contains partial information about a common observation target. The AoC measures the time elapsed since the generation of the latest set of update packets received at the destination. Hence, unlike the age of information (AoI) metric, the AoC decreases only when all update packets from multiple sources for a common observation are collected (i.e., a successful joint update). When simultaneous transmissions from the direct and relay-aided sensors cause packet collisions at the relay, conventional multiuser decoding (MUD) is usually used to decode native packets explicitly from the superimposed signals. Nevertheless, MUD does not work well when the signal-to-noise ratios (SNRs) of different sensors are (nearly) equal. To this end, this paper puts forth a physical-layer network coding (PNC)-aided CDRT scheme for joint information updating, utilizing both MUD and PNC decoders. The PNC decoder decodes superimposed signals into network-coded packets, particularly effective when the SNRs of different sensors are close. We design an automatic repeat request (ARQ) protocol tailored to low AoC and study how network-coded packets can be utilized to reduce the AoC of uplink CDRT, where the closed-form peak AoC formula is derived. We evaluate the PNC-aided CDRT scheme using software-defined radios. Experimental results show that our PNC-aided CDRT scheme significantly reduces the average peak AoC under various SNR conditions.
»Cite this article: Meng, G., Liu, H., & Chan, T. T. (2024, June). Peak age of collection in coordinated direct and relay transmission with physical-layer network coding. In 2024 IEEE 99th Vehicular Technology Conference (VTC2024-Spring) (pp. 1-6). IEEE.
Paper Title:
Effective Search Strategy for Moving Targets in Unknown Environments Using Multiple Robots
Authors:
Ye, S., Chang, M. Y., Chan, T. T., Liu, H., Wang, Y., & Yu, L.
Abstract:
Robots are widely used for target search in applications such as search and rescue, environmental monitoring, and surveillance. Existing search algorithms typically rely on target signals or predictable movement patterns, which might not be available in real applications. In this paper, we address the practical challenge of a target search problem where targets do not emit signals and have unpredictable movement patterns. The problem is formulated as an area coverage problem: how to maximize the coverage of the robots’ detection areas within a limited time. We propose an algorithm that divides the search area into multiple partitions and assigns specific partitions to robots for maximizing their coverage and the success rate of target detection. Within each partition, the random walk technique is adopted by the robots to handle robot failures and unknown obstacles. Through theoretical analysis and experiments, we explore the optimal number of partitions as well as the optimal partition shape to facilitate searching. Extensive simulations across different dynamic environments validate the effectiveness and adaptability of our proposed algorithm.
»Cite this article: Ye, S., Chang, M. Y., Chan, T. T., Liu, H., Wang, Y., & Yu, L. (2024, October). Effective search strategy for moving targets in unknown environments using multiple robots. In 2024 IEEE 30th International Conference on Parallel and Distributed Systems (ICPADS) (pp. 76-83). IEEE.
2023
Paper Title:
Design of Closed-Loop Cold Chain Logistics Optimization Model
Authors:
Lam, H. Y., Tang, V., & Ho, G. T.
Abstract:
Recently, the demand for cold chain logistics has significantly increased due to increased awareness of the safety in frozen foods, improved quality standards for luxury goods, and also the advancements in pharmaceutical technology. Especially with the impact of COVID-19 leading the sharp increase in demand for vaccines, cold chain logistics has begun to be emphasized in recent years as it must consider issues such as passive and active packaging, as well as environmental concerns. Most importantly, logistics vehicle route optimization is a critical factor that can significantly impact business profitability. This paper aims to design a closed-loop cold chain logistics model (CCCL), to minimize unnecessary costs and improve business competitiveness. The model takes environmental issues into consideration, as protecting the environment is a necessity for modern businesses. The CCCL model is capable of handling reverse logistics and the use of recycled packaging materials, which reduces costs while meeting environmental requirements. Simulation result is presented to illustrate how the CCCL model can be applied in planning and pickup and delivery in a cold chain.
»Cite this article: Lam, H. Y., Tang, V., & Ho, G. T. (2023, December). Design of Closed-Loop Cold Chain Logistics Optimization Model. In 2023 IEEE International Conference on Industrial Engineering and Engineering Management (IEEM) (pp. 0315-0319). IEEE.
Paper Title:
A Blockchain-based Decision Support System for E-commerce Order Prediction
Authors:
Ho, G. T., Tang, Y. M., Lam, H. Y., & Tang, V.
Abstract:
The rapidly growing e-commerce sector has created new opportunities and challenges to the logistics industry. Nonetheless, the majority of Hong Kong logistics industry, especially small-and-medium (SMEs) firms, lack operational decision support to adequately handle the seasonal, fragmented, and fluctuating e-commerce orders. To grasp the opportunities of the e-commerce logistics business, logistics service providers (LSPs) should enhance their capability in information exchange and operational planning. With these improvements, the logistics industry would be better able to sustain and expand their e-commerce logistics business. In this paper, a Blockchain-based E-Commerce Analytics Model is developed to enhance digital supply chain integration. Firstly, timely operational decision support would be achieved as blockchain technology, after that, the ML algorithm would enable the logistics industry to manage data efficiently and to forecast dynamic e-commerce order demand. Subsequently, the proposed model allows LSPs to flexibly re-allocate the right number of resources in real time to deal with the hour-to-hour fluctuating arrival of orders in distribution centers. Additionally, the proposed model enables logistics practitioners to predict the sales performance related to e-commerce.
»Cite this article: Ho, G. T., Tang, Y. M., Lam, H. Y., & Tang, V. (2023, February). A blockchain-based decision support system for E-commerce order prediction. In 2023 International Conference on Artificial Intelligence in Information and Communication (ICAIIC) (pp. 041-045). IEEE.
Paper Title:
Study on Innovation Resistence for the New Product Development of a Smart Wine Cabinet Solution
Authors:
Wu, C. H., Tsang, Y. P., & Hui, W.
Abstract:
This studyanalyses the viticulture sector’s evolution, emphasizing the need for a novel wine management system. A surge in wine consumption during the 20th century, attributed to lifestyle changes and rising disposable incomes, has necessitated more accessible and efficient management systems. However, existing wine-centric mobile applications and even premium wine cabinets lack comprehensive wine management features, making tracking and locating specific bottles challenging. We propose the ‘Wise Wine’ system, comprising a Smart Wine Cabinet and a Smart Wine Mobile Application. The system offers standard wine cabinet functionalities plus a wine recognition function linked to the mobile application, enabling real-time tracking of stored wines. The study develops and examines the measures for evaluating innovation resistance concerning Wise Wine’s new product development, resulting in an effective sales and marketing strategy.
»Cite this article: Wu, C. H., Tsang, Y. P., & Hui, W. (2023, August). Study on Innovation Resistence for the New Product Development of a Smart Wine Cabinet Solution. In GSRD International Conference (pp. 34-40). Institute for Technology and Research (ITRESEARCH).
»ISBN (Print): 978-93-90150-28-1
Paper Title:
Digital Twin based Packet Reception Prediction for C-V2X Networks
Authors:
Hou, Y., Zhang, Z., Li, W., Fan, M. H., & Lam, C.
Abstract:
C-V2X (Cellular Vehicle-to-Everything) is a wireless communication technology that facilitates communication between vehicles, infrastructures, and pedestrians via cellular networks, with the aim of improving road safety and efficiency. However, in autonomous mode, transmitting vehicles are unable to estimate the intended broadcast packet reception, resulting in the use of blind broadcast. This limits the reliability and latency performance of C-V2X and hinders transmission scheme optimization. To address this issue, we propose a simple but accurate prediction model to estimates packet reception rate at the transmitter side without requiring additional signaling from the receiver side. Our proposed model achieves an R2 of up to 0.92 when trained and evaluated by data generated by digital twins for the same city, and an R2 of 0.90 when transferring the trained model from a different city.
»Cite this article: Hou, Y., Zhang, Z., Li, W., Fan, M. H., & Lam, C. (2023, October). Digital Twin based Packet Reception Prediction for C-V2X Networks. In 2023 IEEE 98th Vehicular Technology Conference (VTC2023-Fall) (pp. 1-6). IEEE.
Paper Title:
Work in Progress: An AI-Assisted Metaverse for Computer Science Education
Authors:
Ho, K. H., Hou, Y., Chu, C. F. C., Chan, C. K., Pan, H., & Chan, T. T.
Abstract:
This paper proposes the use of metaverse for computer science education by leveraging virtual reality (VR) and artificial intelligence (AI) technologies. VR glasses are used to record videos such as presentation slides and classroom lectures in the metaverse. AI technologies automatically generate customized class notes summarized from the video recordings. This education metaverse can accommodate students’ digital avatars in different virtual workspaces, including collaboration spaces, laboratories, and classrooms. It consists of practical features that mimic a physical classroom setup, such as real-time voice chat, avatar movement, and shared screen presentations. In the metaverse journey, the AI notes generator module is used to create personalized class notes using technologies including optical character recognition, automatic speech recognition, natural language processing, and text-to-speech. We discuss current progress and future work to prototype an AI-assisted metaverse for proof-of-concept experiments in computer science education.
»Cite this article: Ho, K. H., Hou, Y., Chu, C. F. C., Chan, C. K., Pan, H., & Chan, T. T. (2023, March). Work in progress: An AI-assisted metaverse for computer science education. In 2023 IEEE World Engineering Education Conference (EDUNINE) (pp. 1-4). IEEE.