The Research Of Optimized Coll-Aborative Filtering Recommendation Technology Based On Hybrid Algorithm In Cloud Computing

The rapid development of information technology has led to the explosive growth of global data,and the beginning of the big data era is also accompanied by the phenomenon of "information overload".Among the many recommendation algorithms,with robustness and high-efficiency collaborative filtering recommendation technology effectively alleviates the problems caused by information overload.However,the technology also faces a few challenges in the application domain need to be solved urgently.In this thesis,the problem of cold start and scalability of the collaborative filtering recommendation algorithm will be analyzed and in-depth studied.First,in this thesis a hybrid algorithm is used to construct the clustering model in collaborative filtering algorithm in term of the cold start problem,with the hybrid algorithm of K-means clustering and improved genetic algorithm is used to obtain the K value of the clustering algorithm and the initial center set adaptively,the contour coefficient is used as the fitness function of genetic algorithm,according to the entropy of the new user or item attribute information to classify,implementing the similarity calculation and nearest neighbor search in the corresponding clustering model,to complete the evaluation and prediction of new users or new items from the evaluation information of nearest neighbor sets,and to realize the final recommendations of the new users or new items.Secondly,the clustering algorithm can alleviate the issue of algorithm scalability to some degree.On this basis,the paper further studies and analysis the specific processing steps of collaborative filtering recommendation algorithm,combining with powerful cloud computing platform in today's information technology,the most widely used MapReduce distributed framework is used to complete the parallel processing of the algorithm,in order to further improve the collaborative filtering recommendation algorithm to deal with scalability issues.Finally,The Iris data set and glass data set of UCI data set are used to verify the rationality of the user and item clustering model.In addition,a series of experiments are carried out on the MovieLens data set in this paper.The validity of the algorithm is given by comparing the proposed algorithm and the MAE value of the traditional algorithm in dealing with the cold start problem,and the algorithm presented in this thesis runs on the traditional stand-alone environment and Hadoop cluster,through the analysis of the results of the algorithm,the efficiency of Hadoop clustering algorithm for parallel processing is verified.