Ensuring a high degree of customer satisfaction is critical to the success of telecom companies in maintaining or increasing market share. Using churn prediction as an indicator of customer satisfaction, we present a robust system that allows telecom companies to gauge the perceived quality of service delivery. The system employs a novel approach known as the churn drift detection method (CDDM), which can effectively predict customer churn by simultaneously addressing the interdependent problems of concept drift and class imbalance in raw data. We have applied CDDM to the call detail record (CDR) of a south Asian telecom company for churn prediction. By measuring criteria like prediction error and detection delay, experiments have shown that on average, CDDM gives 3 percent less prediction error, 50 percent less detection delay, and almost 1-2 percent better evaluation time than existing techniques.

中文翻译：

---

确保最佳电信服务交付的稳健系统方法

确保高度的客户满意度对于电信公司保持或增加市场份额的成功至关重要。使用流失预测作为客户满意度的指标，我们提出了一个强大的系统，允许电信公司衡量感知的服务交付质量。该系统采用了一种被称为流失漂移检测方法（CDDM）的新方法，它可以通过同时解决原始数据中概念漂移和类别不平衡的相互依赖问题来有效地预测客户流失。我们已将 CDDM 应用于一家南亚电信公司的呼叫详细记录 (CDR) 以进行客户流失预测。通过测量预测误差和检测延迟等标准，实验表明，CDDM 平均减少 3% 的预测误差，减少 50% 的检测延迟，