The Research Of Recommendation System Based On Cloud Computing

With the rapid development of mobile Internet, smart phones and tablets becomemore and more popular in people’s daily life, the Apps which are running on theintelligent terminal device have greatly enriched people’s life. People can search anddownload Apps in application distribution markets, but these markets lack effectivepersonalized recommendation, generally recommend several popular Apps to users.People are difficult to find the Apps that are really suitable for their own interests andhobbies in the mass of Apps, but easier to find public taste Apps. And many newApps of high quality are difficult to be found by people, which may reduce the Appdevelopers’ enthusiasm and creativity.This thesis aims to puts forward a solution of the personalized Apprecommendation for intelligent terminal device based on the research of cloudcomputing, Hadoop distributed framework, recommendation algorithms and thecharacteristics of personalized App recommendation, and finally build out therecommendation system.The main research of this thesis contains cloud computing, the Hadoopdistributed architecture, the recommendation algorithm, the improved algorithm forApp recommendation, and the client, server, PCs cluster of the recommendationsystem. The user information collection is based on a desktop App, which runs onmobile phone, pad, and intelligent TV. The system uses BAE cloud server to controleach function module in the background. The recommend computing environment islocal PC cluster with Hadoop framework. The recommendation algorithm is based onMahout user-based collaborative filtering algorithm, and is improved by using theclassification of Apps.This thesis will complete the following work:1) Collect user App usage information. Select the distributed App personalizedrecommendation as the key technical point.2) Analysis of user’s App information and the App’s attributes. Select theclassification of Apps as the main way to improving the recommendationalgorithm. And improve the algorithm by different aspects, such as to lowerdata sparse, similarity calculation, recommendation list and the problem of new users.3) Research on the method of App classification. Put forward the method andstrategy for App classification.4) Build a recommendation system, that contains the BAE cloud server, localPCs cluster and the desktop application to collect user information.