Scholarlink

In the telecommunications industry, a large customer base generates huge amounts of data every day. In this case, acquiring new customers is more expensive than keeping existing ones, and attrition refers to the process of transferring customers from one company to another within a given prescribed time period. With the increasing number of telecom operators, the forecast of telecom customer churn has become an important demand. However, due to the large amount of data, sparse and uneven, telecom customer churn forecasting has always been a complex task. In this paper, Genetic Algorithm (GA) is used for feature selection, and Random Forest (RF), Decision Tree (DT) and other classifiers are used for classification. The experiment was validated on Orange Telecom’s customer churn dataset. And the experimental results show that the RF model based on GA feature selection has the best processing effect on customer churn data, and the accuracy of this model is 0.9535.

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

Enhanced Telecom Customer Churn Prediction Through Genetic Algorithm-Based Feature Selection and Machine Learning Techniques

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Zou Yuxin,
  • Ghulam Mohi-ud-din,
  • Deng Zhiwei,
  • Xiong Yang,
  • Xia Ling Lin,
  • Chen Ai,
  • Hu Min

摘要

In the telecommunications industry, a large customer base generates huge amounts of data every day. In this case, acquiring new customers is more expensive than keeping existing ones, and attrition refers to the process of transferring customers from one company to another within a given prescribed time period. With the increasing number of telecom operators, the forecast of telecom customer churn has become an important demand. However, due to the large amount of data, sparse and uneven, telecom customer churn forecasting has always been a complex task. In this paper, Genetic Algorithm (GA) is used for feature selection, and Random Forest (RF), Decision Tree (DT) and other classifiers are used for classification. The experiment was validated on Orange Telecom’s customer churn dataset. And the experimental results show that the RF model based on GA feature selection has the best processing effect on customer churn data, and the accuracy of this model is 0.9535.