Research On Box Office Prediction Based On Microblog Data

With the rapid development of social networks, more and more researchers begin to focus on how to use social media data to predict the social event in the real world. Social media contains massive data related to user’s thoughts and behaviors. Digging the relevant information and applying it to predict terrorist activities, national elections, and other hot issues of social emergencies is becoming a hot topic of current research which can guide government action, business activity, production and life. In order to learn the feasibility and accuracy that social media predict real events, we use microblog data to predict box office as an exampleIn this thesis, we introduce the research background and significance of the box office predictions, and analysis research status from both early prediction and real-time prediction of box office. In subsequent chapters, we describe the history and characteristics of microblog as well as microblog data grabing and cleaning. Then we propose a feature fusion prediction model based on microblog emotional intensity、promotional features and many other features. In order to verify the accuracy of the model, we introduce three prediction models :linear model, neural network model, support vector machine model.The main contents and innovations of this thesis are as follows:1. An algorithms is proposed to remove the zombie powder and advertisers algorithms which can reduce abnormal data;2. An user behavior analysis method based on emotional intensity is proposed with which can determine the degree of user preferences for a particular movie more accurately;3. After the analysis and classification of microblog, we set the propaganda action of the investors, cinema, actor as an important reference factor to be considered in the predictive model;4. A multi-feature fusion prediction model based on quantitative characteristics, emotional strength characteristics and promotional features is proposed;5. For the junk data appeared in microblog, an improved BP neural network algorithm is proposed with which can improve fault tolerance of prediction model.In the experimental simulation stage, the influencing factors of the box office are analyzed to identify their relevance and compare the accuracy of different forecasting models. Experiments demonstrate the superiority and accuracy of the proposed prediction models and come up with a series of important conclusions, which can provide theoretical guidance for filming, screening schedule and early propaganda.