Movie Box Office Prediction at the Distribution Stage Using Text Mining of Movie Reviews

Most movie box office predictive models use various variables related to movie production, audience, and movie screening. However, one of the most important variables in the distribution stage when a movie is released is the movie review related to the audience, that is the subject of word of mouth. In most predictive models in which various variables have been used so far, movie reviews have been substituted into emotional indices for additional analysis. In this case, it can be said that it has a limitation in that it is difficult to specifically grasp the contents of word of mouth about the movie. In this paper, only words constituting movie reviews on online portal sites were used as independent variables in the movie box office predictive model, and the relationship with movie box office performance was identified using the LASSO model. Through this, the predictive power of movie reviews on movie box office success was evaluated, and the words in movie reviews that affect movie box office were investigated. As a result of the analysis, the LASSO regression coefficients of words had a correlation with the success of the movies. The LASSO regression coefficients increased as the words appeared in reviews of successful movies, but decreased as they appeared in reviews of unsuccessful movies. As a result, it can be said that movie box office prediction using movie reviews can be helpful for planning additional investment strategy at the distribution stage.