Research On Hybrid Recommendation Algorithm Based On Heuristic Similarity And Trust Measure

With the rapid development of information technology and Internet,people have from lack of information era turned into a new era of information overload.People enjoy the convenience brought by the Internet,and are also faced with the problem of finding information of interest in massive amounts of information.Personalized recommendation establishes a preference model through analyzing users' historical behavior,and achieves active recommendation under the situation of information overload.However,personalized recommendation still faces many challenges,which makes it difficult for recommender system to recommend satisfactory information to users quickly and accurately.This study focuses on data sparsity,cold start and trust issues in collaborative filtering recommendation algorithm.The main work is as follows:Firstly,aiming at the problem of data sparsity existing in collaborative filtering algorithm,a recommendation algorithm based on heuristic similarity is proposed.By analyzing the intrinsic relationship between users' score data,three similarity factors are defined.And the PSO algorithm is used to get the optimal weight combination,and then get the heuristic similarity.Then,HS is regarded as a new standard for the search for neighbor users in collaborative filtering.It not only makes the similarity measurement more accurate,but also maximizes the utilization of data in sparse situations.In the generation of recommended lists,a improved KNN recommendation algorithm based on HS and K-means is proposed to generate better recommendation lists.The method divides clustering based on HS and searches for nearest neighbor in user clustering.It not only reduces the search time of the nearest neighbor,but also constructs a more accurate neighbor set.The experimental results show that this algorithm is superior to the traditional collaborative filtering algorithm in alleviating the data sparsity,and can help to improve the quality of the recommendation.Secondly,in order to further solve the problem of data sparseness,cold start and trust,a recommendation algorithm based on transmitted trust is proposed.By analyzing the influence of social network,user trust network is constructed based on the theory of trust propagation.On this basis,considering the indegree and outdegree of trust network and transmission characteristics of the trust,the transmitted trust is defined,which not only can measure the trust relationship between users more accurately,but also can reduce the sparsity of trust data.Then a new similarity measure method HST is obtained by weighted fusion of transmitted trust and heuristic similarity,and use the improved KNN algorithm based on HST and K-means to generate recommendation list.Experimental results show that this algorithm alleviates the data sparsity,cold start and trust problems in collaborative filtering recommendation to a certain extent.Finally,on the basis of the above two algorithms,a recommendation algorithm based on differentiated relationship is proposed.Through the analysis of the relationship between the common scoring items and all the scoring items,and the influence on the user's similarity,differentiated relationship is defined,which makes the user relationship refined.On this basis,the similarity measure HSTR,which combines similarity degree,trust degree and relation degree,is proposed and a recommendation list is generated by using the improved KNN algorithm based on HSTR and k-means.Experimental results show that the algorithm proposed in this paper can measure user similarity more accurately,and the user's preference prediction is more accurate and the recommendation is more personalized.