AI-based Repeat Buyers Prediction System using Deep Learning

Accurate prediction of repeat buyers can effectively help e-commerce companies adjust their business strategies accordingly. This study proposes an AI-based predictive model decision system for identifying. In the system, a complete machine learning and deep learning model comparison and automatic selection mechanism are established through the analysis of performance parameters and time efficiency. This study proposes an AI-based predictive model decision system, in which Support Vector Machine (SVM), Random Forest, Light Gradient Boosting Machine (LightGBM), Light Gradient Boosting and Factorization Machine (LightGBM+FM), and Deep Factorization Machine (DeepFM) are performed in order to determine the optimal predictive model for each e-commerce platform. Based on the parameters of model performance and time efficiency, including accuracy, Area under the ROC Curve (AUC score), Log Loss score, model fitting time and prediction time, the best models for Taobao and Tmall platforms are tested to be the Deep Factorization Machine with a 76.17% accuracy and a 0.7615 AUC score, and Light Gradient Boosting and Factorization Machine with a 61.90% accuracy and a 0.6063 AUC score. The results indicate that this model decision system can help build reliable and economical model strategies for industrial companies based on the repurchase prediction of their e-commerce platforms.