High Utility Item Set Mining and Product Recommendation Based on Customer Purchasing Behaviour Analysis Using Hybrid Ensemble Model

Abstract

In the modern era, high utility itemset (HUIS) mining is constantly gaining widespread popularity owing to its applicability in multifarious arenas such as web service, retail marketing, and many more. HUIS mining provides a means to discover such itemset within a transaction database that has high utility i.e., offers a higher profit. There have been discovered multifarious approaches in recent years for HUIS mining as well as product recommendations. However, the existing HUIS mining models have certain limitations such as lower prediction accuracy, more time ingesting, massive computational complexity, and expensive implementation, etc. To resolve such key problems, this paper presents a novel system architecture based on hybrid ensemble models for HUIS mining and product recommendation based on customer purchasing behaviour analysis in real time. This proposed system architecture has been trained and validated on widely recognized datasets namely the Chainstore and Foodmart. This proposed hybrid ensemble model provides optimal performance metrics i.e., the accuracy of 98.94%, precision of 97.59%, recall of 97.99%, and F1 score of 98.27%, respectively. Though, in recent years, diversified models have been developed for HUIS mining and customer behaviours analytics in real time. However, there is a wind-ranging scope of further research to explore novel and improved models for HUIS mining in a faster manner for analysis of customer purchasing behaviour.