Churn Prediction in Over-The-Top (OTT) for Customer Retention using Machine Learning Algorithms

Abstract

OTT services have transformed the entertainment industry by giving customers rapid access to various content. However, OTT platforms are plagued with churn or subscriber cancellations. This study investigates machine learning (ML) as a churn-reduction strategy. It investigates how machine learning (ML) algorithms might reduce churn and improve customer retention in the OTT arena. We study the usefulness of ensemble learning, Long Short-Term Memory (LSTM) networks, and Multilayer Perceptron (MLPs) in detecting users on the verge of leaving. To solve intrinsic data imbalances in which churned users comprise a smaller minority, we employ SMOTE (Synthetic Minority Over-sampling Tech- nique). We investigate the impact of customer experience, content library, and consumption patterns on turnover likelihood. Our findings reveal that Gradi- ent Boosting (Accuracy: 87.2%), Random Forest (Accuracy: 91.2%), and MLP (Mean Accuracy: 87.1%) achieve strong performance, with SMOTE improving XGBoost’s recall by 697% (0.083 → 0.697), with LSTMs showing even more potential with additional data. This study highlights the potential of ML for OTT platforms to get important consumer information, design focused retention campaigns and eventually reduce churn. This results in long-term financial stability and a devoted subscription base.