Research On Variable-Granularity Content’s Recommendation System Based On User Interest Feedback

With the upgrading of Internet technology and rapid development of smart mobile devices, the convenience of web services in the mobile terminal such as mobile phones and pads are catching up with that in PC terminal. Traditional Internet companies are developing mobile terminal Apps to consolidate and expand the user market. For example, Tencent launched WeChat mobile terminal application to create another mobile social network platform of one hundred million user level, while Alibaba launched mobile phone PayPal to complete 197 million transactions’ payment on November 11th 2014.More and more Internet Companies are entering the mobile Internet field, developing many more application services into App Stores.Users living in the era of mobile Internet have been benefited from the large mobile application services. On the other hand they are facing new confusion called information overloading. Getting real needs and making right choices cost more and more troubles, especially in the field of online shopping, users often have to browse a large number of boring product information before choosing the right goods. The fact that bad experience continuous for a long time may sometimes leads to the decrease of application users. Therefore, it is necessary to introduce effective personalized recommendation technology in the development of mobile Internet.On the basis of studying the advantages and disadvantages of current various algorithms and the special application background of mobile Internet, a variable-granularity content’s recommendation system based on user interest feedback is proposed in the thesis. The main work of this thesis is as follows:1. With studying research status of traditional recommendation technologies, the trend of the mobile Internet era, as well as the special nature of the mobile devices and situation, a variable-granularity content’s recommendation system based on user interest in real-time feedback is designed in the thesis.2. The system is divided into two subsystems, the server subsystem and mobile client subsystem. The subsystem in the server is responsible for generating the candidate recommended-item sets, available mobile-used information by increasing the two-orientation filtering and pushing them to the mobile client according to user double interest model and the subsystem in mobile client is responsible for real-time response to user behavior and mobile situation, dynamically selecting items presented to the user and feeding back the updated user interest model to the server in the appropriate time.3. In this thesis, the latest data-set MovieLens is used to carry out the relevant experiments, and the accuracy, coverage and response time on the performance of this algorithm are compared with that of other mainstream recommendation algorithm in different scale data. The results show that this proposed system has a better performance and can be competent at real-time recommendation of mobile client...