Accuracy Measure of Customer Churn Prediction in Telecom Industry using Adaboost over K Nearest Neighbor Algorithm

Abstract

Aim: To enhance and predict the accuracy rate of customer churning in the telecommunication industry using Adaboost algorithm over K Nearest Neighbor algorithm. Materials and methods: Both Adaboost algorithm and K Nearest Neighbor algorithm with a sample size of (N=10) is executed with multiple training along with testing splits for predicting the accuracy of customer churning shows g-power as 75% and threshold value as 0.000 and confidence interval as 95%. The performance of these algorithms are calculated based on the rate of accuracy using customer churn dataset. Results and Discussion: The accuracy of predicting customer churn using Adaboost algorithm(90%) and K Nearest Neighbor algorithm(75%) is obtained. There was an analytical difference between Adaboost and K Nearest Neighbor (p<0.005). Conclusion: Prediction of customer churn using Adaboost algorithm seems to be comparatively better than the K-Nearest-Neighbor algorithm with improved accuracy.