Predicting Movie Ratings with Machine Learning Algorithms

The fact that a film is a hedonic product makes it difficult to assess its quality before consumption, therefore consumers who want to reduce uncertainty need various quality signals in their decision-making processes. In recent years, adding to movie-related information, user reviews or ratings on online movie databases have become important quality signals, where many movie viewers use these sites to decide which movie to watch or whether or not to watch a certain movie. In this study, it is attempted to estimate the rating and popularity of a movie by using the main product features as the origin, production year, actor and plot. A database containing 8943 movies shot between 2000 and 2019 from the website sinemalar.com is formed with the help of a web crawler Latent Dirichlet allocation topic extraction is applied to plots and assigned topics obtained from LDA analyzes, along with other movie-related attributes are used to predict the rating class and popularity class of a movie by employing machine learning algorithms such as random forest, gradient boosting tree and decision tree. Using the random forest algorithm attribute statistics, based on their contribution to the predictive power of the model the relative variable importance is also examined.