Customer Retention Modeling over the OTT Platform using Machine Learning

Objectives: Customer retention, a multifaceted issue that plagues the digital entertainment industry, particularly within the realm of Over-the-Top (OTT) platforms, poses a significant challenge, impacting revenue and sustainability. With the discontinuation of subscriptions or usage, retention not only impacts immediate revenue streams but also threatens the long-term sustainability of these platforms. Recognizing this challenge, this paper undertakes a comprehensive examination of predictive modeling methods tailored explicitly for forecasting customer retention within OTT platforms. Method: To construct predictive models, a diverse array of datasets is harnessed, encompassing a wide spectrum of customer-related variables. These datasets include demographic information, viewing history, subscription patterns, and various engagement metrics. Leveraging these datasets, advanced machine learning algorithms are deployed to develop robust models capable of predicting customer retention. Findings: The scrupulous study evaluates the implementation of a range of machine learning algorithms, including logistic regression, random forests, AdaBoost classifier, Decision Tree, and K nearest neighbor (KNN) classifier. Assessment metrics such as F1-score, recall, precision, and accuracy are employed to show the effectiveness of the employed models for customer retention modeling within OTT platforms. The results reveal that the highest accuracy of 80.40% is obtained using the AdaBoost classifier. Novelty: The research uses attribute significance analysis as a means of identifying the fundamental factors that impact client retention. OTT providers can receive important insights into the elements causing subscriber attrition by identifying these main drivers, which will help them develop tailored retention strategies. Keywords: Neural networks, Logistic regression, Random forests, Decision trees, KNN, AdaBoost classifier, Retention prediction