Personalized Recommendation System For Aggregated Information Based On Social TV

The booming of Internet has brought much pressure to the development of traditional TV industry, but taking advantage of the tendency of Internet+, it will have chance to glow the new vitality. Social TV is the opportunity to save traditional TV industry. Due to the seal of itself and lack of interaction between users, traditional TV industry is losing a large number of audiences. With the rise of new media and new ways such as combination of different screens, traditional TV is easy to find a breakthrough point. Social TV is aimed at adding social media and mobile applications to traditional TV industry, thus producing large numbers of user generated content (UGC) and huge amounts of relevant information. However, facing the vast amounts of data, users are getting lost. On the one hand, users don't know how to follow the information.On the other hand, critical users want to enjoy the unique data services. In this context, considering that different people have different needs, the fans group is chosen as our group to study and serve because of their potential of generating abundant UGC. The thesis is intended to provide personalized recommendation of Social TV aggregate information for fans group in this thesis.First of all, this thesis starts with the obtain of information. The main resource of information is web crawler, and UGC is taken into consideration as well. However, the information is of varying quality. On this occasion, data cleaning and structuring storage are very important,thus to provide guaranteed basis for recommendation in the next step.Secondly, we study personalized recommendation algorithm based on user behavior. Considering the particularity of fans group, star property is the essence of aggregate information and what we focus on.Based on user behavior data gathered by our mobile application in the system, interest models are established for different users. Faced with the problem of cold start and real-time problem, we take users' similarity into account as bias term to improve the availability of the system.At last, personalized recommendation system is designed and implemented, in the form of Android Application. To support the function of APP, the server takes the responsibility of database management,running recommendation algorithm and so on. The application is mainly used to present recommendation result and collecting user behavior.Implementation of various modules belong to the system will be introduced in detail in the thesis.