In many applications such as dynamic social network and customer behavioral analysis, the data in... more In many applications such as dynamic social network and customer behavioral analysis, the data intrinsically have many dimensions and can be naturally represented as high-order tensors. In this study, a SVM ensemble learning method is proposed for classification using tensor data. The method is used in identifying cross selling opportunities to recommend personalized products and services to customers. Two real-world databases are used to evaluate the performance of the method. Computational results show that the SVM ensemble learning method has good performance on these databases.
Cross-selling is an integral component of customer relationship management. Using relevant inform... more Cross-selling is an integral component of customer relationship management. Using relevant information to improve customer response rate is a challenging task in cross-selling. Incorporating multitype multiway customer behavioral, including related product, similar customer and historical promotion, data into cross-selling models is helpful in improving the classification performance. Customer behavioral data can be represented by multiple high-order tensors. Most existing supervised tensor learning methods cannot directly deal with heterogeneous and sparse multiway data in cross selling. In this study, two novel ensemble learning methods, multiple kernel support tensor machine (MK-STM) and multiple support vector machine ensemble (M-SVM-E), are proposed for crossselling using multitype multiway data. The MK-STM and the M-SVM-E can also perform feature selections from large sparse multitype multiway data. Based on these two methods, collaborative and non-collaborative ensemble learning frameworks are developed. In these frameworks, many existing classification and ensemble methods can be combined for classification using multitype multiway data. Computational experiments are conducted on two databases extracted from open access databases. The experimental results show that the MK-STM exhibits the best performance and has better performance than existing supervised tensor learning methods.
This study develops machine learning methods for the data-driven demand estimation and assortment... more This study develops machine learning methods for the data-driven demand estimation and assortment planning problem by addressing three subproblems, that is, demand forecasting simultaneously considering cross-selling and substitutions, estimation of the cross-selling and substitution effects, and assortment optimization. These three subproblems are transformed into three sequentially related machine learning problems: collective demand forecasting, demand inference for cross-selling and substitutions, and assortment rule mining. For collective demand forecasting, related product features are introduced to consider both the cross-selling and substitution effects, and a collaborative coordinate descent method with a good convergence property is developed to make distributed demand forecasting and a global update of related product features. Using the results, demand inference adopts transfer and semisupervised learning methods to tackle the challenge of missing data in quantifying the cross-selling and substitution effects. For assortment rule mining, the assortment rules bridge the gap between prediction and optimization, and the developed heuristics obtain the best assortment using the prior knowledge discovered in demand inference. The computational results on a real-world database and a semisynthetic database show that collective demand forecasting obtained far better results than the standard demand forecasting methods and some popular graph learning methods, and the developed heuristics identified much better assortments than those obtained with the baseline methods. History: Accepted by Ram Ramesh, Area Editor for Data Science and Machine Learning. Funding: This work was supported by the construction base project of discipline innovation and talent introduction plan of Chinese higher educational institutions (111 project) [Grant B16009] and the National Natural Science Foundation of China [Grant 72031002]. Supplemental Material: The online appendices are available at https://doi.org/10.1287/ijoc.2022.1251 .
Abstract A dual-channel supply chain is considered in which a manufacturer sells its product thro... more Abstract A dual-channel supply chain is considered in which a manufacturer sells its product through a traditional retailer and an e-tailer, wherein the e-tailer may adopt two, i.e., the conventional batch ordering and the drop-shipping, order fulfillment policies. The problem is modeled as Stackelberg games in which the manufacturer is the leader and the retailer and the e-tailer are the followers competing on price. Optimization models are developed and equilibrium solutions are obtained for unknown distributions of the uncertain demands through a distribution-free approach. Analytical solutions of pricing and ordering decisions are also derived when the uncertain demand is assumed to follow known distributions to evaluate the performance of the distribution-free approach. Numerical studies are conducted to evaluate the effectiveness of the distribution-free approach and sensitivity analyses are performed to examine the effects of several important parameters on the equilibrium solutions. The choice of the e-tailer and the preference of the manufacturer for the order fulfillment policies are also analyzed through numerical studies when important parameters change. The results show that the e-tailer prefers the batch ordering policy if its market share is relatively large and its profit-sharing ratio is relatively low, the manufacturer prefers the batch ordering policy if the e-tailer’s market share is relatively small and the e-tailer’s profit-sharing ratio is relatively high, and both the e-tailer and the manufacturer prefer the drop-shipping policy for other situations. The drop-shipping policy benefits the manufacturer and the e-tailer when the channel substitutability is relatively low and the demand fluctuations are relatively high.
A combined location-routing-inventory system (CLRIS) in a three-echelon supply chain network is s... more A combined location-routing-inventory system (CLRIS) in a three-echelon supply chain network is studied with environmental considerations. Specifically, a bi-objective mixed integer programming model is formulated for the CLRIS to deal with the trade-offs between the total cost and the carbon-capped difference (CCD). A multi-objective particle swarm optimization (MOPSO) heuristic solution procedure is developed and implemented to solve the bi-objective mixed integer programming problem. The bi-objective mixed integer programming model and the MOPSO heuristic procedure are applied to a real-life problem as an illustrative example. The approximate nondominated frontier formed by solutions not dominated by others can be used for the decision makers to make trade-offs between the total cost and the CCD. Sensitivity analyses are conducted, and the relationship among the carbon cap, CCD, the total cost and the carbon prices are examined, and relevant managerial insights are provided. Comp...
Cell-like P systems with promoters/inhibitors (PIC-P systems) are a class of distributed and para... more Cell-like P systems with promoters/inhibitors (PIC-P systems) are a class of distributed and parallel computing models inspired by the function of regulating biochemical reactions by enzymes in biological cells. In PIC-P systems, each evolution rule (rule for short) can be associated with one promoter/inhibitor which helps control the evolution process of objects. In this work, we propose a novel variant of PIC-P systems, called time-free cell-like P systems with multi-promoters/inhibitors (time-free MPIC-P systems for short). In such systems, each rule can be associated with multiple promoters/inhibitors, and the execution time of each rule is extended from one time unit to a random number of time units. These two characters make the time-free MPIC-P systems closer to biological cells. We also investigate the computational power of time-free MPIC-P systems. As results, it is achieved by simulating the matrix grammar that such systems can generate the set of lengths of recursively e...
Data and file type classification research conducted over the past ten to fifteen years has been ... more Data and file type classification research conducted over the past ten to fifteen years has been dominated by competing experiments that only vary the number of classes, types of classes, machine learning technique and input vector. There has been surprisingly little innovation on fundamental approaches to data and file type classification. This chapter focuses on the empirical testing of a hypothesized, two-level hierarchical classification model and the empirical derivation and testing of several alternative classification models. Comparative evaluations are conducted on ten classification models to identify a final winning, two-level classification model consisting of five classes and 52 lower-level data and file types. Experimental results demonstrate that the approach leads to very good class-level classification performance, improved classification performance for data and file types without high entropy (e.g., compressed and encrypted data) and reasonably-equivalent classific...
Abstract This study provides a systematic overview of the literature in gray market business usin... more Abstract This study provides a systematic overview of the literature in gray market business using a data-driven approach and points out several future research directions. The emergence of gray markets is discussed first and the research method and literature selection used in the study are then introduced. By studying the selected articles from five databases, the research trend, highly productive scholars and frequently cited papers are listed next using the collected dataset. Furthermore, frequencies of keywords used in the selected literature are presented through word segmentation and descriptive statistics. Using a data-driven approach, this study divides the extant literature into four categories, including pricing, distribution channels, impacts on the brand-name product manufacturer/equity and consumer attitude, and others. Finally, because the previous studies focused less on information, data, technology and human behaviors, this work puts forward five research agendas and proposes future research directions, including cooperation-oriented, information-oriented, behavior-oriented, technology-oriented, and research method-oriented.
In many applications such as dynamic social network and customer behavioral analysis, the data in... more In many applications such as dynamic social network and customer behavioral analysis, the data intrinsically have many dimensions and can be naturally represented as high-order tensors. In this study, a SVM ensemble learning method is proposed for classification using tensor data. The method is used in identifying cross selling opportunities to recommend personalized products and services to customers. Two real-world databases are used to evaluate the performance of the method. Computational results show that the SVM ensemble learning method has good performance on these databases.
Cross-selling is an integral component of customer relationship management. Using relevant inform... more Cross-selling is an integral component of customer relationship management. Using relevant information to improve customer response rate is a challenging task in cross-selling. Incorporating multitype multiway customer behavioral, including related product, similar customer and historical promotion, data into cross-selling models is helpful in improving the classification performance. Customer behavioral data can be represented by multiple high-order tensors. Most existing supervised tensor learning methods cannot directly deal with heterogeneous and sparse multiway data in cross selling. In this study, two novel ensemble learning methods, multiple kernel support tensor machine (MK-STM) and multiple support vector machine ensemble (M-SVM-E), are proposed for crossselling using multitype multiway data. The MK-STM and the M-SVM-E can also perform feature selections from large sparse multitype multiway data. Based on these two methods, collaborative and non-collaborative ensemble learning frameworks are developed. In these frameworks, many existing classification and ensemble methods can be combined for classification using multitype multiway data. Computational experiments are conducted on two databases extracted from open access databases. The experimental results show that the MK-STM exhibits the best performance and has better performance than existing supervised tensor learning methods.
This study develops machine learning methods for the data-driven demand estimation and assortment... more This study develops machine learning methods for the data-driven demand estimation and assortment planning problem by addressing three subproblems, that is, demand forecasting simultaneously considering cross-selling and substitutions, estimation of the cross-selling and substitution effects, and assortment optimization. These three subproblems are transformed into three sequentially related machine learning problems: collective demand forecasting, demand inference for cross-selling and substitutions, and assortment rule mining. For collective demand forecasting, related product features are introduced to consider both the cross-selling and substitution effects, and a collaborative coordinate descent method with a good convergence property is developed to make distributed demand forecasting and a global update of related product features. Using the results, demand inference adopts transfer and semisupervised learning methods to tackle the challenge of missing data in quantifying the cross-selling and substitution effects. For assortment rule mining, the assortment rules bridge the gap between prediction and optimization, and the developed heuristics obtain the best assortment using the prior knowledge discovered in demand inference. The computational results on a real-world database and a semisynthetic database show that collective demand forecasting obtained far better results than the standard demand forecasting methods and some popular graph learning methods, and the developed heuristics identified much better assortments than those obtained with the baseline methods. History: Accepted by Ram Ramesh, Area Editor for Data Science and Machine Learning. Funding: This work was supported by the construction base project of discipline innovation and talent introduction plan of Chinese higher educational institutions (111 project) [Grant B16009] and the National Natural Science Foundation of China [Grant 72031002]. Supplemental Material: The online appendices are available at https://doi.org/10.1287/ijoc.2022.1251 .
Abstract A dual-channel supply chain is considered in which a manufacturer sells its product thro... more Abstract A dual-channel supply chain is considered in which a manufacturer sells its product through a traditional retailer and an e-tailer, wherein the e-tailer may adopt two, i.e., the conventional batch ordering and the drop-shipping, order fulfillment policies. The problem is modeled as Stackelberg games in which the manufacturer is the leader and the retailer and the e-tailer are the followers competing on price. Optimization models are developed and equilibrium solutions are obtained for unknown distributions of the uncertain demands through a distribution-free approach. Analytical solutions of pricing and ordering decisions are also derived when the uncertain demand is assumed to follow known distributions to evaluate the performance of the distribution-free approach. Numerical studies are conducted to evaluate the effectiveness of the distribution-free approach and sensitivity analyses are performed to examine the effects of several important parameters on the equilibrium solutions. The choice of the e-tailer and the preference of the manufacturer for the order fulfillment policies are also analyzed through numerical studies when important parameters change. The results show that the e-tailer prefers the batch ordering policy if its market share is relatively large and its profit-sharing ratio is relatively low, the manufacturer prefers the batch ordering policy if the e-tailer’s market share is relatively small and the e-tailer’s profit-sharing ratio is relatively high, and both the e-tailer and the manufacturer prefer the drop-shipping policy for other situations. The drop-shipping policy benefits the manufacturer and the e-tailer when the channel substitutability is relatively low and the demand fluctuations are relatively high.
A combined location-routing-inventory system (CLRIS) in a three-echelon supply chain network is s... more A combined location-routing-inventory system (CLRIS) in a three-echelon supply chain network is studied with environmental considerations. Specifically, a bi-objective mixed integer programming model is formulated for the CLRIS to deal with the trade-offs between the total cost and the carbon-capped difference (CCD). A multi-objective particle swarm optimization (MOPSO) heuristic solution procedure is developed and implemented to solve the bi-objective mixed integer programming problem. The bi-objective mixed integer programming model and the MOPSO heuristic procedure are applied to a real-life problem as an illustrative example. The approximate nondominated frontier formed by solutions not dominated by others can be used for the decision makers to make trade-offs between the total cost and the CCD. Sensitivity analyses are conducted, and the relationship among the carbon cap, CCD, the total cost and the carbon prices are examined, and relevant managerial insights are provided. Comp...
Cell-like P systems with promoters/inhibitors (PIC-P systems) are a class of distributed and para... more Cell-like P systems with promoters/inhibitors (PIC-P systems) are a class of distributed and parallel computing models inspired by the function of regulating biochemical reactions by enzymes in biological cells. In PIC-P systems, each evolution rule (rule for short) can be associated with one promoter/inhibitor which helps control the evolution process of objects. In this work, we propose a novel variant of PIC-P systems, called time-free cell-like P systems with multi-promoters/inhibitors (time-free MPIC-P systems for short). In such systems, each rule can be associated with multiple promoters/inhibitors, and the execution time of each rule is extended from one time unit to a random number of time units. These two characters make the time-free MPIC-P systems closer to biological cells. We also investigate the computational power of time-free MPIC-P systems. As results, it is achieved by simulating the matrix grammar that such systems can generate the set of lengths of recursively e...
Data and file type classification research conducted over the past ten to fifteen years has been ... more Data and file type classification research conducted over the past ten to fifteen years has been dominated by competing experiments that only vary the number of classes, types of classes, machine learning technique and input vector. There has been surprisingly little innovation on fundamental approaches to data and file type classification. This chapter focuses on the empirical testing of a hypothesized, two-level hierarchical classification model and the empirical derivation and testing of several alternative classification models. Comparative evaluations are conducted on ten classification models to identify a final winning, two-level classification model consisting of five classes and 52 lower-level data and file types. Experimental results demonstrate that the approach leads to very good class-level classification performance, improved classification performance for data and file types without high entropy (e.g., compressed and encrypted data) and reasonably-equivalent classific...
Abstract This study provides a systematic overview of the literature in gray market business usin... more Abstract This study provides a systematic overview of the literature in gray market business using a data-driven approach and points out several future research directions. The emergence of gray markets is discussed first and the research method and literature selection used in the study are then introduced. By studying the selected articles from five databases, the research trend, highly productive scholars and frequently cited papers are listed next using the collected dataset. Furthermore, frequencies of keywords used in the selected literature are presented through word segmentation and descriptive statistics. Using a data-driven approach, this study divides the extant literature into four categories, including pricing, distribution channels, impacts on the brand-name product manufacturer/equity and consumer attitude, and others. Finally, because the previous studies focused less on information, data, technology and human behaviors, this work puts forward five research agendas and proposes future research directions, including cooperation-oriented, information-oriented, behavior-oriented, technology-oriented, and research method-oriented.
Uploads
Papers by Minghe Sun