Customer Churn Prediction Using Tunicate Swarm Optimization Based Hybrid Machine Learning Algorithms

The customer churn is referred as number of previous customers who leads to leave the service provider through particular time. The main goal is to predict churn customers as early stage and find the cause for its churn. This paper proposed a Tunicate Swarm Algorithm-Support Vector Machine-Improved Random Forest (TSA-SVM-IRF) for predicting customer churn. The IBM Telco dataset is used in this paper and it is preprocessed by one hot encoding. At this time, the data is converted into numerical form. The TSA is used for feature selection which overcomes dimensionality detection and vanishing gradient issues. Then, the SVM-IRF is used for predicting churn customers at early stage. The f1-score, recall, precision, and accuracy are used to estimate the model performance. TSA-SVM-IRF attains 97.59% f1-score, 98.86% recall, 96.64% precision and 96.82% accuracy when compared to existing techniques like Grey Wolf Optimization and Ensemble Neural Network (GWO-ENN), and Swish Recurrent Neural Network (S-RNN).