Movie success prediction with statistical analysis techniques and machine learning methods

In the movie industry, huge investments have been made to shoot a successful motion picture. However, despite large investments, there are some movie examples that cannot be successful as expected. Therefore, predicting the success of a movie is so important on a large scale for the movie producers before releasing the movie. In this study, a classification-based prediction model is aimed to develop for providing a foresight to the producers about investing on a movie. Different statistical analysis and machine learning approaches were used in the proposed model for predicting success of a movie. We mainly focus on detecting which movie attribute is highly correlated with the success of the movie and which machine learning technique is better at predicting the movie success. To do so, firstly a statistical analysis was conducted by using chi-square analysis and analysis of variance test. Then a comparative analysis was performed by using different machine learning techniques including Random Forest, Support Vector Machine and Artificial Neural Network, and Multiple Linear Regression, Support Vector Regression and Random Forest Regression methods as regression methods. The experimental results indicate that the most important predictors of a movie's success are "voteAverage", "voteCount", "revenue" and "budget. In addition to this, Random Forest has become successful by the accuracy of 96% in predicting movie success among other machine learning methods, and Support Vector Regression has become successful by the Root Mean Square Error (RMSE) 1.77 in predicting movie success among other regression methods.