The Research And Implementation Of Micro-blog Recommendation System Based On Cloud Computing

At present, with the development of Internet and the popularity of network applications andmobile applications, there are more and more users of micro-blog platform and they areproducing large amounts of data. As the amount of data increases, the traditional data processingdifficult to meet people’s needs, users need faster and more efficient data processing. For user notsatisfied with the status quo to promote the development of cloud computing andrecommendation systems.In order to solve the above problems, the main job is as follows:1.the paper briefly introduces the concept, deployment patterns, SPI service model and thecurrent research status at home and abroad. Then focus on the analysis of the Apache communityorganizations to provide open source cloud computing Hadoop parallel programming frameworksystem. And by actively trying, paper successfully builds the Hadoop clusters which provides theinfrastructure protection for the following experiments.2.this paper introduces the related technology of recommendation system. Focus on theanalysis of the recommendation process, characteristics and key technologies of differentrecommendation algorithms. Focuses on the analysis of the User-based collaborative filteringand Item-based collaborative filtering. And Comparison of the present various algorithms tosolve the cold start problem. Pointed out the advantages and disadvantages of each method.Facing the problem of sparse data, this paper proposes a hybrid collaborative filteringrecommendation algorithm. User-based Collaborative filtering is very dependent on thesimilarity between users, and Item-based Collaborative filtering is very dependent on thesimilarity between items. Hybrid collaborative filtering recommendation algorithm. It combinestwo kinds of collaborative filtering algorithm, avoid excessive reliance on user or itemproperties.3.This paper analyzes the characteristics of micro-blog platform features and micro-blogusers. The user is divided into: new users and old users. According to the type selection of therecommended way. New users according to their custom labels, select the tag recommendationalgorithm; old users according to their attribute selection recommendation algorithm. Thenumber of following is more than fans, selection of User-based collaborative filteringrecommendation algorithm; the fans more than the number of following, selection of Item-basedcollaborative filtering recommendation algorithm.4.Through the experiment, verify the superiority of the hybrid collaborative filteringrecommendation algorithm. In the recommendation quality, the hybrid recommendationalgorithm is compared with the User-based collaborative filtering recommendation algorithm andItem-based collaborative filtering recommendation algorithm. Experiments show that the higherquality of hybrid recommendation algorithm. In response time, to prove the superiority of the algorithm from two aspects of data size and number of nodes.5.the hybrid collaborative filtering recommendation algorithm combined with Hadoop. Thedesign and implementation of a data set to Sina micro-blog recommendation system prototype.The system uses Hadoop high operation speed, storage capacity and other advantages of theinfinite, optimization recommendation algorithm. Makes the recommendation system deal withthe huge amount of data can be fast, efficient. The recommendation system in parallel computing,fault tolerance and scalability are improved.