Predicting Over the Top Services Movies and Shows Success Using Machine Learning

Product innovation plays a crucial role in the film business, especially in the highly competitive and uncertain landscape of the entertainment sector. To stay ahead of the competition, filmmakers, investors, and OTT services need to continuously innovate and create successful movies and shows. The film business is a multi-billion-dollar entertainment sector with significant hazards. The excessive costs of production still do not guarantee the success of a movie or show. To overcome this uncertainty, machine learning technology can be used to predict the success of a film. Currently, people are starting to watch movies and TV shows on the over-the-top (OTT) services. Netflix is an OTT service that has a lot of subscribers. With such a large business opportunity in this industry, it would be better if the stakeholders could predict the success of a film earlier, so filmmakers, investors, and OTT services will benefit from it, such as by saving costs. Therefore, this study's purpose is to find the best model for predicting the success of Netflix movies and shows using various attributes, namely type, release year, age certification, runtime, genre, production country, IMDB vote, TMDB popularity, TMDB score, and Top Movie (The class label generated from IMDB score, which classifies a movie to be successful (“Yes”) if the score is ≥ 8 and unsuccessful (“No”) if the score is <8). This paper applied Random Forest, Naive Bayes, and k-Nearest Neighbors machine learning algorithms for predicting the success of a movie or show. This study shows the comparison of the measurements of the three algorithms without and with SMOTE. The best algorithm in this case is Random Forest with 81.59% Accuracy, 87.41% Precision, 73.82% Recall, 91.30% AUC, and 79.48% F1-Score.