Recommendation Algorithm Design Based On Collaborative Filtering

Because of its rapid development and universal application,information technology provides convenience for daily life,but also brings "user data troubles".How to filter redundant information in the face of massive data and efficiently provide users with personalized needs Effective information has become a major problem that needs to be overcome urgently.Therefore,the recommendation system technology came into being and became an effective means to solve this problem.The collaborative filtering algorithm has become one of the most widely used algorithms in the recommendation system due to its unrestricted content and recommendation objects,less user interference,and easy implementation of technology.The traditional user-based collaborative filtering algorithm has many problems,such as time-consuming search for users’ near neighboors,low quality of near neighbors,and lack of analysis of users’ individual difference needs.This thesis mainly focuses on how to optimize the recommendation effect of user-based collaborative filtering algorithm.By dividing the target users into communities,perfecting the search method for the set of nearby users,improving the similarity formula of users and revising the project prediction score,high-quality recommendation results can be obtained.The main work of this thesis is as follows:(1)Community division of user collections.Community segmentation of users’attribute information by fuzzy clustering algorithm,determine the core user and cluster center of each community,and lock the search range of the nearest neighbor user,so as to alleviate the complexity of the original algorithm which needs to traverse all users.(2)Modify Pearson similarity formula.This factor,by increasing the interaction between the user information to define the similarity between the user,in originally only consider "quality" of similarity between the user on the basis of increased to investigate the"quantity",will be based on user ratings behavior originally recorded similarity measure method is extended to comprehensive consideration of user attributes and historical action,Increases recommendation accuracy and interpretability for user-based recommendations.(3)Define user execution and project popularity.It is used to modify the scores of items in the initial recommendation list,reorder the list,relieve the limitation of the recommendation results,meet the personalized different needs of users based on the common preferences.(4)Improve the recommendation algorithm.On the basis of the above work,the traditional user-based collaborative filtering recommendation algorithm is improved,and an improved recommendation algorithm(MUCF-R algorithm)is proposed.The experimental results show that the MUCF-R recommendation algorithm is interpretive and accurate in recommendation.The rate has been improved,and it is an effective recommendation algorithm.