Research And Application On Movie Box Office Prediction Oriented On Social Media

With the rapid development of internet,social media becomes the platform for people to share and apply the information.As one of the popular art,movies are becoming one of the special goods on the internet.Movie box office prediction is an important research topic in data mining.It can provide a more comprehensive and reliable suggestion both for publicity and risk assessment,and it can process a better application value in prediction field.This thesis proposes different classification prediction model to achieve movie box office prediction on the base of actors/actress influence in social media.The main research work is shown as follows:1)This thesis combines user basic properties and behavior attribute to build user influence model.By clustering analysis,it can complete the judgment of the performer influence.2)This thesis proposes classification and prediction model for the movie before it releases,then analyzes the actors/actress influence,and then constructs two box office prediction models,support vector machine model and neural network classification model.Achieve box office prediction before the movie releases,by means of the performer influence in social media.3)This thesis proposes a multi-linear regression prediction model for the movie after it releases.The movie classification is done by constructing the user rating matrix and using the latent factor model(LFM).The multi-linear regression model is constructed for different classifications.Experiment result shows that the proposed approach is feasible.The impact of actors/actress can reflect the influence of the actors/actress in social media.The precision of the movie box office prediction model before release can reach above 74%.The precision of movie box office prediction model after release can reach above 90%.At last,we describe the exiting problems and the next research works.