Box Office Prediction Based On Consumer Intention And Sentiment Analysis

With the rapid development of the film industry,the output of films has continuously increased,and the number of movie theaters has also increased.For film investors,the most important thing is the level of the movie box office,because the level of the box office can reflect whether the investment in film produces income.How to use film box office prediction to determine the investment of film for publicity to reduce investment risk has gradually become a key link in film investment.Social media is booming,and the texts related to movies published by users on social media often contain consumer wishes and emotional colors,which makes text mining the most popular topic.The microblog platform has many celebrities and movie stars from all walks of life,and has great influence.Therefore,data mining on movie-related comment data on microblog has great significance.Based on the microblog review text,this article cuts in from three perspectives:intention analysis,sentiment analysis,and microblog popularity to extract relevant features,using Linear Regression,Logistic Regression,Decision Tree,Random Forest Regression and SVR-RBF five algorithms for movie box office prediction.Among them,in the research of microblog heat,this article focuses on the extraction method of topic distribution feature of microblog heat.Based on the traditional topic distribution model,the microblog heat is fused to obtain the topic distribution characteristics of the microblog heat.The experimental results show that the microblog hot topic distribution feature extraction method is superior to the traditional topic distribution feature extraction method in box office prediction.Finally,in the use of multiple algorithms for movie box office prediction,when only a certain feature is used for prediction,the relative error value of the model is large,indicating that the model does not perform well.When each additional feature combination is used for prediction,the relative error value will become smaller,and when the consumption intention feature,emotional tendency feature and microblog hot topic feature are used for prediction at the same time,SVR-RBF performs best and the relative error is the smallest.