Machine Learning–Based Random Forest Prediction of Online Shopping Behavior in the Digital Economy

Against the backdrop of the continuous development of the digital economy, consumers' online shopping behaviors have become increasingly complex. Predicting consumers' purchasing behavior has become a major challenge. Based on the e-commerce behavior data in February 2021, this study constructed a consumer online purchase prediction model using Random Forest. The performance of prediction model for out-of-sample data has been enhanced through rigorous data cleaning, feature engineering, and class balancing. The accuracy and recall exceed the 0.90, and the F1-score exceeds 0.80. Exploratory data analysis has demonstrated to reveal behavioral patterns related to time and price. Furthermore, this study has proposed significant strategies for the enhancement of conversion of shopping carts into purchases. Feature importance analysis in this study has found that the four important factors influencing purchase intention are as follows: cart_event_ratio, total_events, cart_per_product and view_count. This study proposes a practical, data-driven framework for e-commerce practitioners to predict consumers' online purchasing behavior more accurately. Furthermore, the conversion strategies for online shopping will be effectively improved in the highly competitive digital economy.