Abstract: To address the issue of intelligent package recommendation for telecom users, a prediction method based on user group segmentation was proposed. In this approach, the K-means clustering algorithm was first applied to classify users into five distinct groups, considering their behavioral characteristics. Within each group, the XGBoost model was then utilized to predict package selection. The dataset included multidimensional features such as user demographics, package details, consumption behavior, usage patterns, and service quality. Experimental results demonstrated that the proposed method significantly improved prediction accuracy compared to traditional regression models, with an accuracy increase ranging from 0.176 to 0.217, reaching a prediction accuracy of 92.8%. These findings indicate that precise user group segmentation and predictive modeling enhance the effectiveness of package

Key words: User group segmentation, Consumer behavior analysis, Data mining, Telecom service optimization, K-means clustering

摘要： 为解决运营商用户智能套餐推荐问题，提出了一种基于用户群体划分的套餐选择预测方法。该方法首先利用K-means聚类算法将用户划分为五个群体，考虑到群体特征的差异性。接着，在每个群体内，基于XGBoost模型进行套餐选择预测。所用数据集包括用户基本信息、套餐类型、消费行为、使用量及服务质量等多维特征。实验结果表明，相比传统的单一回归算法，所提出的方法在套餐选择预测精度上有显著提高，精度提升范围为0.176至0.217，达到92.8%的预测精度。该方法通过精确的用户群体划分和预测模型，能够有效提高套餐选择的预测效果，为运营商优化套餐推荐提供了有力支持。

关键词: 用户群体划分, 消费行为分析, 数据挖掘, 运营商服务优化, K-means聚类