Collaborative Filtering Personalized Recommendation Of Merging Users’ Attributes And Interests Contrast

To solve the problem of information overload and respond to the needs of users’ personalized services, personalized recommendation technology came into being. The paper hopes that users can quickly and accurately find the resources they want through optimization and innovation of personalized recommendation. Among so many personalized recommendation technologies, collaborative filtering algorithm is the hot research currently. Because the range of applications of this algorithm is the widest, the time of development is the longest, and the algorithm is the most sophisticated. Collaborative filtering recommendation is mainly based on users’historical consumption and scoring data which is stored in the database of the system to analyze the user’s interest and predict what products users may consume in the future, thus it implements personalized recommendation to users. Previous studies of collaborative filtering algorithms are mainly based on users’rating matrix to perceive users’preferences and determine the similarity of interests between users by the similarity of users’score. With the development of algorithms, especially there have been algorithm bottlenecks such as scalability problem, cold-start problem which appeared to be inadequate relying solely on the scoring matrix data to find the nearest users. Therefore, we must look for other sources of valid user’ preference data. This paper is to innovate in the user preferences perception, in addition to analyzing traditional users’rating data, but also introducing information of the users’preference attribute which is important sensory data sources, with scoring matrices together constituting the data base of perception of users’ preferences. Users’attributes as describing the users’important information of the individual characteristics, different attributes can be divided into different categories of user groups. Among these users, there may be some interest preference similarity, finding out common interests of these specific user groups to produce the recommendation foundation. This paper defines the parameters of a new index to measure the preferences of user interest which is interests contrast. On this basis, this paper presents a combination of user attributes and interests contrast collaborative filtering personalized recommendation algorithm. The algorithm is a combination of user attributes as constraints, combined with interesting contrast to be co-produced a set of recommendations, and after finishing deleting selection forms the final list of recommendations. The new algorithm proposed in this paper overcomes the traditional collaborative filtering scalability issues as improvement goals, and does not rely on traditional users’similarity calculation to find the nearest neighbor throughout the design process of the new algorithm. Therefore, when the users and projects are rapidly growing, the complexity of the algorithm does not appear sharply increasing. The experiments show that the collaborative filtering recommendation algorithm of integration of user attributes and interests contrast can to achieve satisfactory recommendation quality under the condition of guaranteeing real-time nature of the recommendation which are flexible and efficient recommendation proposal, but more importantly providing a new recommendation ideas. In addition, this paper also systematically analysis recommendation efficiency under different combinations of attributes, so as to establish a foundation of relevant research in this field.