Research On Movie Box Office Prediction Method Based On Review Text And Neural Network

With the rapid development of China’s economy,movies have gradually become an important part of people’s entertainment and leisure,and have also brought huge economic benefits to the society.However,with the gradual strengthening of the state’s supervision of the film and television industry and the sharp shrinking of capital,film and television companies now pay great attention to the relationship between the initial investment and the final box office income during the entire cycle from the establishment of the film project to the release.Movie box office is of great significance to the film and television industry.Through the research on the existing movie box office prediction models,it is found that there are problems such as insufficient selection of box office influencing factors and traditional prediction models.Therefore,selecting more comprehensive influencing factors and using a more accurate prediction model can improve the accuracy of prediction.This paper is mainly based on the deep learning algorithm to study the movie box office prediction.The research results are as follows:(1)According to the three different stages of project approval,production,and publicity before the movie is released,11 categories including popularity index,movie type,release time,main creator list,first-day box office and a total of 34 influencing factors are selected to predict the movie box office.Based on the domestic mainstream movie data platform,a crawler program is written to obtain data to ensure the accuracy and integrity of the data set.Crawling movie reviews through crawlers,and preprocessing the review data to create a corpus based on the movie domain.Using Word2 vec word vector and SOPMI method,a feature thesaurus and sentiment dictionary based on the film field are established.Using the trained Snow NLP library,the TF-IDF algorithm is used to calculate the sentiment score of movie reviews.(2)A multiple linear regression movie box office prediction model based on review text is designed.The model verifies the validity of the review data for box office prediction,and also reveals the influence of various factors on the movie box office.The first-day box office prediction model is established by using the number of people who want to watch and the popularity index of the film in the early stage,which provides an important basis for the prediction of the total box office in the later period,and also provides decision-making reference for the film and television industry and related management departments.(3)According to the selected influencing factors,a movie box office prediction model based on BP neural network and convolutional neural network is designed.The prediction model is optimized through experiments,and the prediction accuracy is improved.In the research of box office prediction based on convolutional neural network,the review text data is used as feature value for box office prediction,and a box office prediction model based on review text and convolutional neural network is constructed,and the process of data processing in the model is optimized through experiments.,which improves the efficiency and accuracy of the prediction model.Through comparative experiments,it is proved that the prediction effect based on the review text and the convolutional neural network model is better than that of the traditional regression model and the BP neural network model.The prediction accuracy rate reaches 85%,and the optimal prediction accuracy rate is close to 90%.