Research On Movie Box Office Prediction Based On K-means Clustering And BP Neural Network

In recent years,with the rapid development of the national economy,people's living standards continue to improve,entertainment needs are rising.More and more viewers choose to walk into the cinema to experience the visual impact of the film and the spirit of pleasure.The domestic film market speed expansion,attracting the influx of many enterprises in this industry to pursue the big profits.But the film as a cultural product,There are many uncontrollable factors that affect the box office.The investment rate of return is difficult to estimate.So it is very meaningful to predict the box office in front of a movie and help the stakeholders of the film to allocate resources reasonably and reduce the investment risk.Due to the rapid development of the Internet,the network is full of massive data,many of which are related to the film-related data.Using the network data to predict the film box office is a hot research topic in the box office prediction field which combined with information technology field,It is also the starting point of this study.First of all,based on the analysis of a large number of domestic and foreign research,the basic process of forecasting and the conceptual model of box office prediction are put forward.Based on the conceptual model,this paper studies the relevant theories and key technologies needed at box office forecasting,including commonly used forecasting model,data reduction method,clustering algorithm,data acquisition technology,Which provides theoretical and technical guidance for later research.Secondly,based on the analysis of the factors affecting the box office and the characteristics of Sina microblogging,Baidu search engine,Douban movie website,according to the design principle of the index,the box office forecasting index system including 14 second-level indicators is established,and the quantitative method of the index is studied one by one.Thirdly,based on the analysis of the existing problems in the neural network box office forecasting model,the RST-K-means-BP prediction model is proposed.The model integrates three kinds of data processing method: BP neural network,rough set theory and cluster analysis.Firstly,the indicators which are reducted through the rough set theory is used as theneural network input layer node,and then the K-means clustering analysis is used to classify the sample data,so that the similarity of the samples in the same class is higher,finally,BP neural network model is established for different categories to forecast.This combination eliminates the redundant attributes,simplifies the neural network structure,improves the training speed of the network and the prediction accuracy of the model.Finally,the RST-K-means-BP prediction model is analysed by using 200 real data obtained from the network.The experimental results show that the RST-K-means-BP prediction model has a large degree of improvement in the prediction accuracy and efficiency compared with the single BP neural network model.At the same time,the article predicts the final box office of a film that has not been released and suggests a proposal to the related party.The application example shows that this research can provide decision support for the film parties,and has good research foreground and application value in the field of prediction.