Friend Recommendation Feedback Algorithm Based On Clustering And Interest Cognition

In interest-based friend recommendation,it is usually necessary to calculate the similarity between different users for the recommendation.However,when the number of users is very large,it takes a lot of time to calculate the similarity of interests among all users.In order to save time,this paper proposes an MSI clustering algorithm based on the ISODATA algorithm,which can cluster sparse feature samples with low time complexity.At the same time,an algorithm based on interest and cognition is proposed on the basis of tripartite graph diffusion algorithm.In addition,a feedback mechanism is introduced to dynamically adjust the recommendation model.The main work of this paper is as follows:(1)On the basis of ISODATA,an MSI clustering algorithm is proposed,which can cluster sparse multi-dimensional attributes.At the same time,a new method is used to realize clustering splitting and merging operations.The time cost of clustering is also low.The MSI algorithm chooses whether to merge or split according to the number of clustering samples.The algorithm automatically adjusts the number of iterations according to the current state,which makes it easier to use.In order to make it more suitable for friend recommendation scenarios,the steps of soft clustering are added to reduce the probability of losing clustering edge samples in the process of recommendation.(2)Based on the tripartite graph diffusion algorithm,a tripartite graph diffusion algorithm based on interest and cognition is proposed.The new algorithm can make full use of the ternary relationship among users,items,and labels,and puts forward the cognitive degree as a measurement standard,which can consider users behavior more carefully.When calculating similarity,the weight of different items and labels is taken into account,and the recommendation effect of users with sparse data is improved.(3)A feedback mechanism is proposed,which can dynamically adjust the recommendation model according to each recommendation result.The feedback mechanism includes user similarity feedback and weight feedback.User similarity feedback adjusts the similarity matrix between different users according to their friends' preferences.Weight feedback adjusts the mixed weights of interest similarity and cognitive similarity according to users' preferences for interest or cognition.Finally,the overall experimental comparison is made.The first step of this recommendation process is to use MSI algorithm to cluster users.The second step is to use interest-based and cognitive-based tripartite graph diffusion algorithm to calculate the similarity between users for the recommendation.The last step is to use the feedback mechanism to adjust the recommendation model according to the recommendation results.Experiments show that this algorithm can reduce the time cost and improve the recommendation effect.