Research On Movie Box Office Prediction Based On Random Forest

Movie marketing is an indispensable factor in the promotion of movie box office,and active marketing strategies can increase movie box office income.After the creation of the movie,it is necessary to improve the marketing and increase the box office revenue as much as possible.This paper starts from the marketing factors affecting movie box office,selects eight variables,such as schedule,premiere box office,and Baidu Index,and uses the data of 173 domestic movies from 2014 to 2016 as research objects.Using the random forest method to build a movie box office regression prediction and classification forecasting model,forecasting and analyzing the box office of 12 movies in 2017,combining two methods to reasonably predict the box office.At the same time,it analyzes the marketing factors that affect the movie box office and proposes a series of suggestions for cinema marketing.Random forest is a commonly used classification method,it can be applied to various kinds of regression problems including numerical,metric and survival variables.Random forests have been extensively studied for regression and classification.However,when the response variables contain order information,there is no perfect targeted solution.This article focuses on the situation that the response variables in the movie box office classification forecast contain ordinal information,and innovatively introduces the random forest model based on conditional inference tree to classify the movie box office.The results of the research are as follows:1.The classification and forecast results of the box office show that the forecast accuracy of the random forest model based on the conditional inference tree is higher than that of the traditional random forest model.The classification prediction results for the 12 movies in 2017 show that the forecasting accuracy of the random forest model based on the conditional inference tree is better than that of the traditional random forest model.2.The regression forecast results at the box office show that the traditional random forest model is more effective than the linear regression model,and the prediction error of most movies is within 30%.3.While using the traditional random forest method to build the box office forecasting model,giving the order of the importance of a variable.The results show that the first day box office and the average person-time of each movie have the greatest impact on the movie box office.Baidu Index and Point to box office in the second place,the Bean score and Sina Weibo have little effect,and the impact of the schedule and domestic movie protection month is negligible.