Intelligent classification method for e-commerce customers based on improved XGboost model

Abstract

To improve the intelligent management of e-commerce customer data, this study proposes an intelligent classification method for e-commerce customers based on an improved extreme gradient boosting tree model. The method uses an improved particle swarm optimization algorithm to optimize the hyperparameters of the extreme gradient boosting tree, so as to improve the model’s ability to adapt to complex data. The outcomes indicated that the proposed model achieved 91.6% in terms of accuracy, which was a 5.0% improvement over the baseline model. The recall rate was 88.9% and the F1 value was 90.2%. In addition, the proposed method was applied to an e-commerce platform. The obtained customer categorization accuracy was 92.3%, the recall rate was 92.1%, the processing cost was 0.8 yuan/thousand times, and the processing efficiency was 480,000 items/day. In terms of robustness, the accuracy using the proposed method was stable at around 90%, with AUC fluctuations ranging from 0.91 to 0.95, and the maximum memory fluctuation of 15.6%. The CPU utilization rate always remained within the safety threshold of 92%. In terms of real-time, the average response time was 28 ms for a single request, 210 ms for batch processing, and 41 ms under peak traffic. The CPU utilization in the three cases was 22%, 48%, and 61%, respectively. The study’s findings demonstrate that the suggested approach can offer e-commerce clients an accurate and effective intelligent categorization solution.