Research On Intelligent Recommendation Algorithm Based On Clustering

With the rapid development of electricity providers to enhance economic efficiency,various problems of recommended technology has gradually highlighted.Collaborative filtering algorithms,as one of the most commonly used techniques of recommendation,its problems that sparsely populated,difficult to expand and others are great challenges for today's large-scale recommended systems.Therefore,researching on collaborative filtering algorithms in this thesis has extremely important value,whether on the basis theory of recommended or in the application practice level.From the theoretical exploration to the experimental verification,domestic and foreign recommended technology research results are comprehensive analysis in this thesis.The idea of innovation is put forward in the research of collaborative filtering algorithm,which integrates the idea of clustering and recommendation technology to alleviate the problem in the recommended technology.The main content and innovations of this thesis include:(1)Attribute Weighted FCM clustering algorithm based on K-means is studied and implemented.The K-means algorithm and the attribute weighting factor are added to the FCM algorithm,which overcomes the shortcomings of the FCM algorithm's dependence on the initial center and disregard of the disconsistent of the attribute.In addition,FCM algorithm has the advantage of dealing with uncertain data sets.This new study maximizes the advantages of FCM clustering,and the experiment also proves the advantages of the algorithm in the thesis(2)FCM clustering algorithm based on W2 VE is studied and implemented.The processing semantic similar strengths of Word2 Vec are used,and it is combined with EMD.A dissimilarity measure based on W2 VE is developed to measure the semantic differences between word eigenvectors.From the level of semantic analysis,the limitations of FCM algorithm to deal with the text was broken,the similarity of the text attributes in the samples are retained on the greatest degree.This is also confirmed by experiment,the new study has advantages in each performance.(3)An intelligent recommendation algorithm based on clustering is studied and implemented.From the perspective of clustering,the problems of collaborative filtering algorithm are solved one by one.The two new clustering algorithms are embedded into the User-based collaborative filtering algorithm and the Item-based collaborative filtering algorithm.Then selection one from the User-based collaborative filtering algorithm and the Item-based collaborative filtering algorithm's results of recommendation using the switching strategy of the minimum recommended error in appropriatly time.And the two modules of algorithm that the new User-based collaborative filtering algorithm and the new Item-based collaborative filtering algorithm,constitute the intelligent recommendation algorithm based on clustering.This study has been proved by experiments,and the new algorithm has a breakthrough in each performance.