Research On User Similarity Recommendation Algorithm Based On Item Weight

With the continuous growth of the Internet data scale,users can not find the resources they need accurately and quickly.In discussing the problem of information overload,the relevant scholars have discussed a variety of coping techniques,among which the recommendation system is a better solution to information overload.Recommendation system is a personalized information service system for users.It can act well as a hub between users and information resources.The recommendation system will first establish the user's personal model for the service object,then analyze and describe the needs of the user's hobby,and finally recommend the corresponding information to the target user under the agreed recommendation plan.Compared with search engine and other information processing tools,recommender system is more intelligent and personalized in information processing.Online e-commerce sites,as well as Internet systems,such as social and video sites,have been used as a recommendation system and play a core role in these network platforms.Among many recommender systems,collaborative filtering recommendation system is a kind of recommendation method which is widely researched by scholars and applied widely.The core of the collaborative filtering recommendation algorithm is to find a project with high score of the nearest neighbor of the target user,and the nearest neighbor is determined mainly by the similarity between the user and the project.The higher the reliability and accuracy of the target user's nearest neighbor is,the better the recommended results will be on the level of quality and benefit,so the success of the recommendation system is based on the similarity calculation method designed with the researchers.This paper introduces the reasons for the formation of the recommendation system and the personalized application of the recommendation algorithm under the current Internet environment.At the same time,we introduce the user behavior,content,near neighbor and situational awareness recommendation system in detail,and explain the respective recommendation principles,so that readers can understand the overall implementation architecture of different recommendation systems.As the core similarity algorithm in the recommendation system,we mainly introduce the implementation of the Pearson correlation coefficient and the angle cosine similarity as the main algorithm in the recommendation system,and give a detailed introduction to the evaluation index of the recommended quality of the recommended system.In the current recommendation algorithm,the proportion of the high and low score projects involved in the calculation is the same,and the recommended target of the recommendation system is to recommend the items that are popular among users,and the weight of the high and low grade project is considered equally,which only reduces the recommendation quality of the recommended tasks and does not get the desired results.On this basis,we propose a user similarity recommendation algorithm based on project weight through consulting relevant information.Through the improvement of user similarity algorithm and the experimental test of MovieLens movie data set,we find that the similarity recommendation algorithm considering the weight of the project is better than the traditional algorithm which does not distinguish between high and low score items,and it can be significantly improved in the accuracy of recommendation,the prediction of the score and the quality of the recommendation.