AI-Powered Customer Purchase Prediction based on Shopping History and Browsing Patterns

This paper presents the use of machine learning (ML) in online buying behavior prediction from user interactions and browsing history with the help of Random Forest algorithm. A large dataset was utilized, and sophisticated preprocessing methods like feature scaling, categorical encoding, and the Synthetic Minority Oversampling Technique (SMOTE) were employed to handle class imbalance and data complexity. The Random Forest model performed superbly with an accuracy of 96.3% and an AUC value of 0.97; it can efficiently classify the purchased and nonpurchase users. Page Values, Product Related Duration, and Bounce Rates are significant predictors of purchase. Actionable insights emerge for targeting strategies in optimizing e-commerce through customer engagement. The paper further delves into ethical considerations about using predictive modeling, such as bias mitigation and compliance to data privacy requirements. With the use of data-driven insights, companies operating in retail, finance, and healthcare can thus improve decision-making, user experiences, and strategic growth. Future research will investigate more alternative models for ML that could include deep learning techniques to further optimize predictive accuracy and scalability.