Research On Influencing Factors Of Box Office Of Domestic Films Based On Supervised Learning

Film is the mainstay of China's cultural industry.It not only enriches people's spiritual and cultural life,but also brings huge benefits to the country's economic development.In 2019,the box office of Chinese films as a whole also exceeded 64 billion yuan,and the development of Chinese film industry is changing from high speed to high quality.However,the film related parties cannot use scientific methods to effectively evaluate the final box office performance of films,and cannot reasonably plan and adjust the operation and management policies according to the gap between the target box office and the actual box office,which will often cause waste of resources and increase the profit risk.Therefore,it is of great practical significance to establish a model to predict the final box office of the movie and to study the influencing factors of the box office of the movie,so as to improve the quality of the movie for the producer,to arrange the distribution marketing strategy reasonably for the distributor,to help the related parties to allocate resources rationally,and to improve the final box office of the movie to a greater extent.In this paper,730 domestic films with the final box office of more than 10 million yuan from 2010 to 2018 were taken as research samples,and 657 of them were randomly selected as model training sets and 73 as model prediction effect test sets.From when the ticket book,optical network,art network,cat's eye network and weibo these five website integrate the corresponding data collection,based on the movie first week of the thinking of related data to predict the final box office,from the aspects of film production,distribution,marketing three built production companies,different types of film,director influence,distribution companies,on the first day box office,the first week of box-office 25 domestic box-office impact factors such as indicator,and the indicators and the relationship between the film grossed a visualization analysis.Use the multiple linear regression,random forests in supervised learning,the Lasso regression and Lasso-XGBoost four models to forecast the divided good test set,from stability,numerical accuracy and classification accuracy prediction effect comparison model prediction effect,verified the Lasso-XGBoost combination model proposed in this paper as a generality and accuracy,the optimal domestic movie box office forecasting model,the numerical prediction accuracy was 72.38%,the classification accuracy is 79.45%,the effect is superior to the commonly used multiple linear regression model and stochastic model of the forest.Finally,according to the optimal Lasso-XGBoost model,the importance order of factors influencing the box office of domestic films was given.Year of release;In terms of marketing,the 11 indicators of first-day box office,first-week box office,first-week score,number of audience wanted to see and number of audience participation in the first week are the important factors to predict the total box office in the first week of release.After a movie's opening week,marketing factors affect the continued growth of the movie's subsequent box office more than any other factor,which in turn affects the performance of the total box office.At the end of this paper,according to the conclusion of the research,it provides feasible suggestions for film production and distribution marketing.