TELECOM CHURN PREDICTION USING MACHINE LEARNING ALGORITHMS
N.GURUMOORTHI
ABSTRACT
Customers are the base for any business success and that is why firms become aware of the significance of acquiring satisfaction of customers. Customer churn is an essential issue and it is regarded as one of the most essential concerns among firms because of increasing rivalry among firms, increased significance of marketing strategies and customers conscious behavior in present years. Organizations must develop different strategies to resolve the churn issues relying on the services they offer. Customer churn practice is essential in competitive and rapidly developing in telecom sector. The process of migrating from one service provider to another telecom service provider occurs due to good services or rates or due to various advantages which the rivalry firm provides customers when signing up. Due to the greater cost related with acquiring new customers the prediction of customer churn has developed as an indispensable part of planning process and strategic decision making in telecom sector. The main aim of the study is to explore the customer churn prediction in telecom using in machine learning algorithms.
In this project Machine learning techniques have been used for estimating the telecom customer probability to churn. This study makes use of Logistic Regression, SVM, Random Forest, ADA Boost and XG boost with big data for predicting consumer churn in the telecom sector. Logistic regression has been used widely to estimate the probability of churn as a function of variables set or features of customers. Support Vector Machine (SVM) has been successfully used in many applications such as image recognition, medical diagnosis, and text analytics. Random forest is a flexible, even without hyper-parameter tuning, a great result most of the time. ADA boost works on the principle of learners growing sequentially. Similarly, XG boost implementation of gradient boosted decision trees designed for speed and performance. This study uses Kaggle’s website dataset for telecom churn to predicting and analyzing churn. The results of the study show that the accuracy rate of prediction in consumer churn is found to be 0.829 percent for a XG Boosting method.
