Prediction Model of User Purchase Behavior Based on Machine Learning

Abstract

In recent years, with the continuous popularity of the Internet, the number of online shopping users in my country has reached 639 million, which contains huge commercial value. In order to maintain the prosperity, diversity and order of merchants, and fully meet consumers' one-stop shopping needs, it is necessary to analyze and predict user purchase behaviors more accurately. This article uses the interaction between users and products within a certain period of time provided by JD.com to predict whether users will place orders in designated categories and shops in the next week. First of all, we aim at the problem that the original features are not strongly related, and use feature fusion to construct strong features. For high-dimensional systems that are slow to calculate and prone to overfitting, we use random forests to filter features, reducing the complexity of the model. Finally, a fusion algorithm based on XGBoost and LightGBM is proposed to predict the purchase behavior of users, and the prediction effects of random forest and GBDT algorithm are compared at the same time. The results show that the fusion model has the best prediction effect, and can help merchants more accurately conduct marketing activities to potential customers based on the prediction results.