Prediction on Customer's Purchasing Decision based on Machine Learning Classification and Regression Models

This study investigates whether users will place an order for a specific item on an e-commerce platform (using JD.com as a case study) based on the comprehensive consideration of all available information. U sing machine learning methods, it analyzes a wide range of related factors to predict purchasing decisions. The dataset is preprocessed and feature-engineered to train predictive models. Key algorithms, including Decision Tree, Random Forest, and Logistic regression, are used for classification tasks, with hyperparameter tuning to ensure optimal performance. In the study, the Random Forest model demonstrates remarkable performance, achieving an accuracy score of 0.999817 and an Area Under the Curve (AUC) of 0.9998, reflecting a high probability of precise predictions. Notably, the features with the most significant predictive contributions are identified as the product's coupon discount level and quantity discount level, collectively accounting for over 50% of the predictive power. This research contributes to the field of e-commerce analyses by offering a data-driven approach to understanding and predicting consumer behavior, with potential applications in personalized recommendations and marketing strategies.