Research on Telecom Customer Churn Prediction Method Based on Data Mining

Aiming at overcoming the shortcomings of common telecommunication customer churn prediction models as single model and poor classification performance, a gradient decision tree integration model (GBDT) prediction model is proposed, and the important parameters are searched by harmonic search algorithm (HS). We built a telecom customer churn prediction model based on HS-GBDT algorithm. This model compares the parameter combinations to be optimized in the GBDT algorithm into the synthesized harmony in the HS algorithm, and seeks the optimal parameter combination of the GBDT model through continuous iteration of the harmony. The experimental results show that the combined model has higher classification accuracy than Logistic regression, support vector machine and random forest, and can provide good decision support for major telecom operators in the process of customer churn management.