Abstract

In this study, we aim to develop a data-driven and personalized retail marketing system based on the relationship between each product and its transaction behaviour. This research will further analyze the correlation between products' transaction histories and customer preferences to refine the personalized retailing system. We segment customer-purchasing behaviour using clustering to enable these characteristics in our retail marketing system. We used correlation analysis and experimental design to engineer and select features. Association rules were used to identify causal associations on a transaction dataset that had already been split based on quartile time horizon and customer-purchasing group. The Network Analysis revealed that the VIPs' transactions were more complex than the Potential Customers', meaning that the VIPs' patterns of purchases were more random, and the Potential Customers were grouped. Potential customer transactions are clustered into small groups, whereas VIP transactions are scattered. In Main Path Analysis, the products' core paths are plotted, and the biggest accumulative weight on the graph is eliminated, along with other insignificant associations. An association's critical path analysis determines the most important path. However, since we use historical data for empirical research, real-world datasets should be used to apply our findings for a comprehensive assessment.In this study, we aim to develop a data-driven and personalized retail marketing system based on the relationship between each product and its transaction behaviour. This research will further analyze the correlation between products' transaction histories and customer preferences to refine the personalized retailing system. We segment customer-purchasing behaviour using clustering to enable these characteristics in our retail marketing system. We used correlation analysis and experimental design to engineer and select features. Association rules were used to identify causal associations on a transaction dataset that had already been split based on quartile time horizon and customer-purchasing group. The Network Analysis revealed that the VIPs' transactions were more complex than the Potential Customers', meaning that the VIPs' patterns of purchases were more random, and the Potential Customers were grouped. Potential customer transactions are clustered into small groups, whereas VIP transactions are scattered. In Main Path Analysis, the products' core paths are plotted, and the biggest accumulative weight on the graph is eliminated, along with other insignificant associations. An association's critical path analysis determines the most important path. However, since we use historical data for empirical research, real-world datasets should be used to apply our findings for a comprehensive assessment.