User Behavior Analysis and Predictive Modelling in E-Commerce Using Association Rule Mining

Abstract

The rapid growth of e-commerce has generated an unprecedented volume of user interaction data, including searches, product views, clicks and transactions. Understanding and predicting user behaviors within this digital ecosystem is vital for improving personalization, enhancing customer satisfaction and sustaining competitiveness. This study explores the use of Association Rule Mining (ARM) as a predictive modelling technique to uncover hidden patterns and correlations in user interactions on e-commerce platforms. By applying ARM and advanced embedding-based mechanisms, the research aims to identify consumer preferences, predict purchasing intentions and optimize recommendation systems. Comparative analysis using benchmark datasets such as RecSys2015 and LastFM demonstrates that embedding vector matching models significantly reduce training time while maintaining prediction accuracy. Findings confirm that ARM-based predictive modelling not only provides actionable insights into user decision-making but also improves personalization strategies and operational efficiency. The study emphasizes the necessity of integrating dynamic variables, continuous monitoring and scalable machine learning approaches in e-commerce predictive systems. These insights are crucial for businesses striving to enhance user engagement, tailor digital experiences and sustain long-term growth in a data-driven marketplace.