Market Basket Analysis Using Community Detection Approach: A Real Case

Abstract

Market basket analysis, as an important analysis tool, helps businesses to improve product promotion and to recommend systems development. The association rules based on the Apriori algorithm is the conventional method of market basket analysis. In this research by modeling transactional data as product network, customer network, and customer-product network through applying a well-known network analysis technique, a number of hidden patterns in the connections among customers and products were detected. The required data is taken from an online retail store in Iran. Transactional data over a period of 6 months with around 320,616 transactions in 2015 was analyzed by the community detection method. Extracted networks from the database depict the complexity of patterns in customer behavior or in the behavior of co-purchased products. Ultimately, findings of this research indicated that the applied method would be more practical and expressive to discover complex relations in comparison to other traditional methods like the association rule mining model. In addition, the findings will be capable enough to be used as the basis of new algorithms for developing the recommender systems.