Research On Personalized Recommendation Algorithm Oriented To Sparse Data

With the rapid development of network technology,Internet resources are more and more abundant.However,there is also a problem of information overload.In the era of big data,information processing is required to be efficient and fast.However,there are still many shortcomings in traditional methods such as search engines.Because of this,personalized recommendation systems are gradually emerging and become effective tools for alleviating information overload and improving decision-making efficiency.Collaborative filtering technology is the main means of personalized recommendation system.The sparse linear method is an efficient collaborative filtering recommendation algorithm proposed in recent years,which attracts a lot of scholars' attention.However,the algorithm has some shortcomings such as sensitive to cold start and sparse data.This paper does the following work for the problems of sparse linear method:1.Aiming at the problem that the similarity between items that can't be scored by the same user can't be calculated in the sparse linear model,a sparse linear algorithm based on similar user sets is proposed by analyzing the recommendation process of sparse linear methods.The concept of similar user sets used to maintain the similarities between items.The algorithm takes into account both the user's interest in the recommended items and the users set's interest in items in the recommendation process.2.In order to further improve the recommendation accuracy of sparse linear models,an algorithm that can integrate users' historical behavior information and social information into the sparse linear model is proposed:SLIM based on social relations.The idea of this algorithm is to guide the recommended learning process by sharing the same implicit features in the scoring matrix and social relation matrix.So that recommendations can take into account the user's historical evaluation information and social information.3.In view of the fact that the current social network-based recommendation algorithms do not utilize the similarity information of users,a recommendation algorithm based on social relationships and user interests is proposed.The algorithm analyzes the social relationship between users from the perspective of the transfer of trust relationship,taking into account the user's own historical rating preference as a basis for accurate recommendation to the user.In addition,the thesis analyzes and contrasts the proposed method on dimension through specific experiments.The experimental results show that the optimized algorithm obviously optimizes the problems of cold start and sparse data,and effectively alleviates the information overload.