The Research On Recommendation Algorithm Based On Project Proportion Factor And Group Contribution

With the continuous development of Information technology,Internet technology plays a significant role in life.While,there is a lot of information in the massive data.Without the help of strong tools,people should spend excessive cost in finding effective information.In the context,the recommendation algorithm comes into being.Because it can accurately analyze the user characteristics,initiatively recommend user interested things and exploit the potential hobbies of user.It is respected among users,and has made certain achievements in many areas.However,due to the changing needs of user,the data sparseness,scalability,cold start and other issues of the traditional recommendation algorithm,there is still much room for improvement.The score data set of recommendation algorithm is too sparse that the number of effective overlapping items is too small,or there is a certain amount of inclusion relationship.And for the user's interests changing over time,it eventually leads to an inaccurate recommendation result.Thus,an effective collaborative filtering recommendation algorithm based on the project proportion factor and the time decay function(ITDCF)is proposed in this thesis,to solve the data sparse and the problem of user's interests change.The project proportion factor approach amplifies the similarity among users,and filters out the more consistent users,to locate the user's interests more accurately.Time decay function analyzes the interest of the neighboring users and predicts the interest points of the target users,which makes the recommendation algorithm have higher pertinence in mining users' potential hobbies.The experiments prove that the personalized recommendation algorithm proposed in this thesis has better results in terms of prediction accuracy and project recommendation quality.As the personalized recommendation ignores the social factors and group characteristics,it cannot integrate the preferences of the group members.As far as possible to meet the needs of group consistency of social activities,more and more scholars start to focus on the group recommendation algorithm.The group recommendation algorithm uses multi-user groups as the recommended target to conduct group discovery,internal preference prediction and preference fusion.In this thesis,we propose Group Recommendation Algorithm based on Item Intersection and Contribution(GRIC),and in this algorithm we utilize a new preference fusion strategy based on group contribution(PFSGC).It is based on the preference of the association degree between each user and the group.This strategy considers the interest of each user in the group comprehensively and reduces the impact of outliers on group recommendation.Through the experimental verification,GRIC algorithm is affected by the number of user features,the user characteristics of species and the number of groups.The experimental results show the better accuracy of the proposed algorithm.