Research And Implementation Of Recommendation Engine Based On The Hadoop And Mahout For Intelligent Terminals Cloud Applications

With the rapid development of mobile communication, especially, in recent years,the common application of3G/4G net and the flourish of cloud computing technology,electric commerce is no longer confined to personal computer in the traditional way.Instead, it has been connected with cloud computing and been applied to intelligentterminal; furthermore, it is the mobile intelligent terminal fueled the rapid developmentof E-commerce. Mobile E-commerce can offer different users’ more unique serviceaccording to their personal demands, which is the most obvious and convincibleadvantage. Meanwhile, overload of information and higher spending ask for betterindividual service. However, for the mid-and small-sized enterprises which develop thecloud application of E-commerce on the basis of intelligent terminal, establishing arecommendationsplatform for E-commerce is of tremendous difficulty in technology.Concerning with this problem, my paper aims at setting up anrecommendationengineplatformfor the E-commerce with intelligent terminal cloud application, and providingrecommendationservice for the cloud application of the intelligence terminal within theplatform.Here, in order to provide recommendationservicefor the cloud application ofintelligent terminal, the paper put forward a model of the recommendationengineplatform on the basis of Hadoop and Mahout. Aiming to solve problems in the loudapplication in intelligence terminal, the model combines Off-line Data Analyzing andOn-line Introduction to realize its goal. The Off-line Data Analyzing part of the model ison the basis of open-source cloud computing platform, Hadoop, which endows theplatformmore accurate computing ability and uses open-source data miningframework,namely Apache Mahout, to support the Off-line Data Analyzing, decreasing thedifficulties of setting up the platformof the Off-line Data Analyzing. The onlinerecommendationpart is applied to analyze users’ actions in real time, that is to say,realizing the goal of introduction immediately. If the parts are appropriately combinedtogether, a better recommendation service will be achieved. Then, furthering theanalysis of the distributing distributing item-based collaborative filtering algorithmwhich Mahout has achieved, studying its demerits and coming up with methodto improve it, and at last, conducting simulation experiment to prove the feasibility ofimprovement.Taking the scheme of setting up the recommendationengine platformforE-commerce with intelligent terminal cloud application as its target, the paper creativelyputs forward a model of recommendationengine platformon the basis of Hadoop andMahout, and conducts an analysis of the distributing collaborative filtering computingand works out solutions and strategies to improve its existing defects. In addition, thepaper has setted up the recommendationengine platformby using Hadoop cloudcomputing and Apache Mahout, and has completed the design and realization of thecomprehensive introductive recommendationengine platform system in accordance withobject-oriented design knowledge. At last, the paper has designed relevant test programsto testify the key function of the introductive recommendationengine platform.