Prediction of Netflix Content Viewing Time Using a Machine Learning Classification Model

The OTT (over-the-top) market has become the center of the media industry, showing rapid growth since COVID-19, and the competitiveness of an OTT platform largely depends on the quality and quantity of its contents. In this environment, this study aims to develop a model that can predict viewing time of each content based on its characteristics, utilizing machine learning techniques and the recent viewership data released by Netflix. This study used content characteristics data collected from IMDB and TMDB as predictor variables, and developed classification models applying Random Forest, Support Vector Machine (SVM), and XGBoost algorithms. The performance of each model was evaluated using Confusion Matrix and the AUC (area under the ROC curve) measure. As a result, all three algorithms showed good performance and XGBoost has demonstrated the best performance. The proposed model is expected to support OTT platforms to make a better decision on purchasing content licences or making original contents by effectively predicting the success or failure of the content in advance.