Classification of Customer Churning based on OTT platform data.

Customer Churning is a serious problem for firms since it reduces revenue and stunts long-term development. As a result, anticipating and reducing customer churn has emerged as a key area for machine learning research. We investigate several machine learning models for forecasting customer churn rates. To pinpoint the elements influencing customer outcomes, we first analyse a real-world dataset of an OTT provider. Next, for the purpose of constructing prediction models, we preprocess the data, conduct exploratory data analysis, and apply a diverse set of machine learning methodologies. These include logistic regression, decision trees, the LGBM Classifier, support vector machines, Random Forest, K-Nearest Neighbour Classifier, XGBoost Classifier, as well as various Bagging and Boosting Methods, respectively. Subsequently, we assess the prediction results using evaluation metrics. Our findings demonstrate that, with an accuracy of 89%, the XGBOOST Classifier algorithm surpasses other algorithms in forecasting customer churn rates. We identify crucial factors, including gender, age, subscription duration, weekly minutes of engagement, video consumption, maximum inactive days, etc. Our research underscores the significance of employing machine learning techniques to predict customer churn rates. This application can aid companies in identifying customers who are at a higher risk of leaving and implementing preventive measures to retain them. Businesses may utilize our insights to create strategies that effectively retain customers and enhance their overall success.