Supply chain management represents one of the most important scientific streams of operations research. The supply of energy, materials, products, and services involves millions of transactions conducted among national and local business enterprises. To deliver efficient and effective support for supply chain design and management, structural analyses and predictive models of customer-supplier relationships are expected to clarify current enterprise business conditions and to help enterprises identify innovative business partners for future success. This article presents the outcomes of a recent structural investigation concerning a supply network in the central area of Japan. We investigated the effectiveness of statistical learning theory to express the individual differences of a supply chain of enterprises within a certain business community using social network analysis. In the experiments, we employ support vector machine to train a customer-supplier relationship model on one of the main communities extracted from a supply network in the central area of Japan. The prediction results reveal an F-value of approximately 70% when the model is built by using network-based features, and an F-value of approximately 77% when the model is built by using attribute-based features. When we build the model based on both, F-values are improved to approximately 82%. The results of this research can help to dispel the implicit design space concerning customer-supplier relationships, which can be explored and refined from detailed topological information provided by network structures rather than from traditional and attribute-related enterprise profiles. We also investigate and discuss differences in the predictive accuracy of the model for different sizes of enterprises and types of business communities.
24 August 2022 in Advances in Data Analysis and Classification
In recent years, network representation learning (NRL) has been increasingly applied into web data analysis, such as video, image and text. Most of NRL methods can widely pursue nodes classification, community detection and link prediction tasks. Due to the nodes in these kinds of networks mostly contain the common attributes and share the same neighbors, we identify them as homogeneous networks, also including real-world networks such as social network, citation network and collaborative network. Therefore, it is nature that nodes tends to connect densely with high similarity. Supply network is one of the most typical collaborative network. It is not a homogeneous network, due to the nodes present two roles - supplier and customer. As the importance of supplier-customer relationships has become increasingly apparent that guides modern research and practice, the main impact of such researches is poured into the field of business management and operation research. However, prior studies have indicated that firms tended to manage their relationships in a more structural and relational approach, no existing literature have applied NRL into predicting business relationships. This paper proposes a novel NRL method and presents two contributions. First, we employ network analysis to extract three centralities, which can represent both local and global context. Second, we also include firm profiles as node contents to train the model by using machine learning techniques. To compare with other state-of-the-art NRL methods, our proposal assesses the concepts surrounding structural and relational characteristics and the extension of firm profiles to supply network, and also represents the infrastructure of the social science of business.
In strategic management of retail industry, the advanced investigation by using radio frequency identification (RFID) technology to capture customers’ in-store behavior has been dramatically attracted scholars and practitioners in past ten years. As a small RFID tag attached to the shopping carts can be recognized as surrogates instead of enumerators to trail the customers, it can provide us an objective and direct perspective to observe and measure the in-store behavior of customers. In this article, we present a study on this new type of in-store behavior data named RFID data, which can improve the understanding of purchase behavior of customers with emphasis on meaningful knowledge via analysis of RFID data. In contrast to prior studies in this research field, this paper has paid special attention to shopping time that customers spent in supermarket (so-called stay time), and presents methodological analysis into two folds. First, we develop a bayesian network (BN) model to combine both of purchase behavior and in-store behavior as features. As BN is a probabilistic graphical model, it can provide an quantitative analysis process of purchase behavior decision over stay time and also allow us to interpret the decision process of purchasing in a much more intuitive measurement. The results show BN has a better accuracy than other typical prediction models (linear discriminant analysis, logistic regression and support vector machine). Second, due to BN can estimate and predict in a nonlinear correlation between purchase intention and stay time, we examine a tedium effect on purchase behavior. During the customers wander in shopping, purchase intention represents a non-monotonic phenomena accounting for the long stay time. Finally, we also investigate the sensitivity and specificity of purchase behavior predicted by our proposal in adjustment of decision threshold and implement several business decision-making implications in actual business situations.
28 November 2017 in The Review of Socionetwork Strategies
Consumer behavior analysis is of great significance to retail merchandise marketing. This article aims to establish a customer classification model that includes the complexity of customer shopping paths. First, we select a target area to capture the customer's movement path data. The path data contains a series of points with (x, y) coordinates. We plot the points into a path map via transforming the pixels from the (x, y) coordinates. In this stage, the points are connected by lines according to the time sequence. Secondly, the box-counting method is used to calculate the fractal dimension of each path map. Thirdly, we considered Gaussian function and distribution similarity to improve k-nearest neighbor (KNN) algorithm. In numerical experiments, we use our improved KNN algorithm to learn a customer classification model based on fractal dimension and stay time. Compared with support vector machine (SVM) and traditional KNN classification models, our improved KNN customer classification model has higher accuracy of 0.925 and higher F1-score of 0.926.