Published Paper Details:

The growth of e-commerce has created large volumes of behavioral and transactional data that can be used to anticipate customer purchasing decisions. This paper presents a practical machine learning framework for customer behavior analysis and purchase prediction using demographic attributes, browsing activity, previous purchase information, and product-related features. The study compares Logistic Regression, Decision Tree, Support Vector Machine, and Random Forest classifiers within a structured pipeline consisting of data cleaning, categorical encoding, normalization, feature selection, model training, and evaluation. The documented experimental implementation uses an 80:20 training-testing split and evaluates models through accuracy, precision, recall, and F1-score. Comparative observations indicate that Random Forest provides the strongest overall balance of predictive capability, robustness, and resistance to overfitting, while Decision Tree offers high interpretability, Logistic Regression provides a computationally efficient baseline, and Support Vector Machine remains competitive for complex decision boundaries. The findings are consistent with published e-commerce studies in which ensemble models achieve strong purchase-intention prediction. The proposed framework can support customer segmentation, personalized marketing, product recommendation, sales forecasting, and more efficient allocation of marketing resources. The paper also identifies data quality, class imbalance, changing customer preferences, and model interpretability as important considerations for deployment.