基于代价敏感逻辑回归的电信客户流失预测研究

Research on Telecom Customer Churn Prediction Based on Cost-Sensitive Logistic Regression

摘要: 针对电信客户流失数据集的多维特征和不均衡问题，本文给出了一种基于代价敏感的逻辑回归的电信客户流失预测模型。通过对不平衡样本集分别采用不同权重调整，将代价敏感学习与传统分类算法相结合，建立基于逻辑回归的电信客户流失预测模型，最后对实际电信客户流失进行验证。通过与其他分类器模型的对比显示此方法在各种评估指标上均有更好的表现，更加符合电信业预测客户流失的实际情况。

Abstract: Aiming at the multi-dimensional characteristics and imbalance of the telecom customer churn dataset, this paper presents a cost-sensitive logistic regression-based prediction model for telecom customer churn. By adjusting the unbalanced sample sets with different weights, combining cost-sensitive learning with traditional classification algorithms, a logistic regression-based tele-communication customer churn prediction model is established, and finally the actual telecommunication customer churn is verified. The comparison with other classifier models shows that this method has better performance in various evaluation indicators, which is more in line with the actual situation of predicting customer churn in the telecom industry.