Research on Customer Churn Detection Based on K-means Clustering and Logistic Regression

Abstract

In the digital age, customer churn has become a key factor restricting the sustainable development of enterprises. Traditional analytical methods that rely on experience-based judgment and simple statistics lack systematicity and accuracy, making it difficult to adapt to complex market environments and customer behavior patterns. This study aims to construct a high-precision and practical customer churn detection model to overcome the limitations of traditional methods. First, customer data is optimized through preprocessing steps such as data cleaning, standardization, and feature engineering. Then, based on the RFM model, the K-means clustering algorithm is used to accurately segment customers, identifying different characteristic groups such as core customers, maintenance customers, and risk customers. Finally, a churn prediction model is constructed by combining the logistic regression algorithm and validated using a real telecommunications customer dataset. The model performance is evaluated using metrics such as accuracy, recall, and F1 score. Experimental results show that the K-means + logistic regression combined model achieves a prediction accuracy of 87.5%, a recall of 82.3%, and an F1 score of 84.8%, significantly improving performance compared to traditional models such as logistic regression and random forest alone. Furthermore, age and cash deposit balance are key characteristics influencing customer churn. This research provides scientific decision support for enterprise customer relationship management, helps formulate differentiated customer retention strategies, and has significant practical application value.