Scholarlink

This study presents an effective approach for predicting e-commerce user purchase behavior and conducting user group clustering by integrating a Long Short-Term Memory (LSTM) network with an improved K-means algorithm. The LSTM model utilizes historical purchase data to predict future user behavior, demonstrating high accuracy and stability in capturing behavioral dynamics. Performance evaluation metrics, including Root Mean Square Error (RMSE) of 15.2 and Mean Absolute Error (MAE) of 10.8, confirm an 85.6% prediction accuracy, highlighting LSTM’s capability in supporting precision marketing strategies. To enhance user segmentation, an improved K-means algorithm is introduced, incorporating the K-means++ initialization method to optimize cluster center selection and a dynamic threshold iteration strategy to boost computational efficiency. Experimental results validate its effectiveness in identifying distinct purchasing patterns among users. This combined approach enables e-commerce platforms to segment users based on purchase frequency and spending behavior, facilitating personalized marketing strategies, optimizing resource allocation, and improving conversion rates.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Research on E-Commerce User Purchase Behavior Clustering and Prediction Based on Long Short-Term Memory Network and Improved K-Means Algorithm

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Tuli Chen,
  • Wennan Wang

摘要

This study presents an effective approach for predicting e-commerce user purchase behavior and conducting user group clustering by integrating a Long Short-Term Memory (LSTM) network with an improved K-means algorithm. The LSTM model utilizes historical purchase data to predict future user behavior, demonstrating high accuracy and stability in capturing behavioral dynamics. Performance evaluation metrics, including Root Mean Square Error (RMSE) of 15.2 and Mean Absolute Error (MAE) of 10.8, confirm an 85.6% prediction accuracy, highlighting LSTM’s capability in supporting precision marketing strategies. To enhance user segmentation, an improved K-means algorithm is introduced, incorporating the K-means++ initialization method to optimize cluster center selection and a dynamic threshold iteration strategy to boost computational efficiency. Experimental results validate its effectiveness in identifying distinct purchasing patterns among users. This combined approach enables e-commerce platforms to segment users based on purchase frequency and spending behavior, facilitating personalized marketing strategies, optimizing resource allocation, and improving conversion rates.