Research On Recommendation Algorithms Based On User Interest

With the development of Internet technology, nowadays society has entered the era of information explosion. One of the hot spots of various types of electronic commerce website is personalized recommendation technology. Personalized recommendation technology is an important information filtration mean, which can mine user's interest preference and provide personalized service for users by target user's interest. This technology can solve the problem of overload information and reduce the time and cost of user search, sorting and analyzing data.One of the key in technologies of recommendation system is to model the user interest, and to design a highly effective recommendation algorithm, the system should mine the user's interest accurately. Because of the user's needs and preferences differing from each other, the same user's needs are often not the same in different situations. It leads to the result that the existing recommendation technology is often plagued by sparse data, user cold start and other issues and it causes that the user experience is poor and the recommended system accuracy is low. To this end, this paper considers the impact of user activity on the similarity calculation and optimizes the Pearson similarity algorithm, and based on the user's interest to improve the traditional collaborative filtering algorithm. The work of this paper is summarized as follows:(1) In this paper, we first introduce the recommended methods used in the three kinds of recommendation techniques, which are collaborative filtering recommendation method, content based recommendation method and hybrid recommendation method. Then the collaborative filtering method is introduced, and several kinds of traditional similarity calculation methods are analyzed, and then some performance evaluation indexes and data sets are introduced;(2) Then we analysis the user data, and sum up the activity will affect the accuracy of the similarity calculation, and then introduce the high frequency items and active users on the impact of similarity calculation, based on the traditional Pearson correlation coefficient algorithm optimization, through offline experiments to verify the improved similarity based collaborative filtering recommendation system to improve the quality of the recommended results;(3) Traditional collaborative filtering technology has some unsolved problems, such as the sparsity of scoring matrix, the accuracy of prediction and so on. Aiming at these problems, based on the optimization of Pearson correlation coefficient calculation method, this paper studies the user's interest, extends the user project matrix, and uses SVD, K-means and other technologies, and proposes a collaborative filtering algorithm based on user interest. Experimental results show that the proposed algorithm can improve the accuracy of prediction, and reduce the impact of data sparsity;(4) We design and implement a recommender system based on Mahout, realizing the application of collaborative filtering algorithm, and finally complete the recommended results show.