Research On Personalized Hybrid Recommendation Algorithm And Its Application

Development of the Internet technology, profoundly changes the world, the amount of data in the Internet increase highly. The sharply increase in the amount of information make people to use information with great convenience, so that people out of the lack of information problems to some extent. At the same time, facing with such a large amount of information, how people can get accurate information about their needs is a confusion, especially in the case of people can not accurately describe the characteristic properties of the required information, the existing search tools seem powerless, this will become increasingly strong sense of distress. Therefore, how to achieve "information to find people," so that the useful information can be automatically rendered in front of people who need it, allowing intelligent system to automatically push information to people, recommendation algorithm and application is one of the hot issues of personalized information services research in the current Internet environment.In this paper, we analysis of the common algorithms, each algorithm has advantages and disadvantages. it is is an effective means to mix several recommendation algorithm according to a certain policy to achieve from each other, so as to obtain a better quality recommendation.Based on the analysis of user preference about item properties shows that different user has a clear distinction preference for item properties. By the statistical analysis of user’s preference for item properties, can determine a user case of item properties preference,then proposed a collaborative filtering recommendation algorithm based on user statistics of preference about item properties (UIPPSCF). Base on the research of this algorithm and experiment,the result compared with the traditional user-based recommendation collaborative filtering recommendation algorithm (UCF). The result shows that, UIPPSCF algorithm has better recommendation accuracy precision rate and recall rate than the UCF algorithm. UIPPSCF algorithm although to some extent improved the performance result recommended, but UIPPSCF itself only consider the user ratings and item properties, such as for film, its properties include comedy, romance, action, etc, without considering user attributes, such as the user’s gender, age, occupation and so on. Thus, in order to further improve the quality of recommendation, this paper from the process of multi-attribute complementary perspective, consider having a complementary feature multiple attributes to recommendation algorithm, and then given based UIPPSCF algorithm and based recommendation algorithm of user attributes (UAB) hybrid recommendation algorithm UIPPSCF_UAB. Base on the research of this algorithm and experiment,The experiment result show that the hybrid recommendation algorithm UIPPSCF_UAB can get more desired result than hybrid algorithm UCF_UAB.On the basis study of the algorithm and simulation, this paper designs and implements a prototype system based on the hybrid recommendation algorithm UIPPSCF_UAB. After the user logs in, the system will provide function which users can choose the number of the recommended movie, according to the needs of users, the corresponding film will recommend to the user.