Machine Learning Implementation for Analyzing Consumer Behavior Patterns in E-Commerce Transactions

Abstract

This research aims to implement and compare the performance of various machine learning algorithms for mapping global e-commerce markets based on sales patterns and consumer behavior. The data used consist of global e-commerce transaction records covering product categories, prices, purchase quantities, consumer regions, and transaction times. The research stages include data preprocessing, classification model implementation, performance evaluation, and result analysis. The evaluated algorithms include Random Forest, Gradient Boosting, Decision Tree, AdaBoost, Linear Regression, and K-Nearest Neighbor (KNN). To address data imbalance, the Synthetic Minority Over-sampling Technique (SMOTE) was applied. The results indicate that the Random Forest model achieved the best performance with an accuracy of 99.87%, followed by Gradient Boosting (99.82%) and Decision Tree (99.51%). The application of SMOTE proved effective in improving class balance and enhancing metrics such as Accuracy, Precision, Recall, and F1-Score. Additionally, feature engineering strengthened the model’s ability to capture complex consumer behavior patterns. Overall, the combination of ensemble algorithms, SMOTE, and feature engineering produced a reliable classification model for analyzing consumer behavior in global e-commerce markets. This study contributes significantly to the development of data-driven digital business strategies and the application of machine learning in the digital economy context.