The Recommendation Algorithm Based On Cloud Computing Research

Nowadays, the rapid development in E-commerce has changed the way people livegreatly, we can get what we want just at home by sitting at the computer or using aterminal which can connect into the Internet. But the explosion of information whichgives us a lot of choices also increases the difficulty of choice. It’s difficult for us tofilter out irrelevant information from the mass of information. The recommendationsystem is born in this background. Its functionality is to dig out useful information forus based on certain knowledge. People develop a great many of different techniques torealize it. Collaborative Filtering (CF) is one of the fastest growing-technique, also themost widely userd in the field of electronic commerce. So this dissertation takes CF asthe key point of recommender system research.At first we propose a method which used to mitigate the negative impact on theperformance caused by data sparsity, based on the research of traditional CF algorithm.Traditional similarity measures are extremely strict from a mathematical point, but lesspractical in the situation when lacking of data. For this reason we propose a data fillingmethod which based on item-similarity so that we can provide enough data toalgorithm.After solving the problem of data, we also propose a imcremental algorithmwhich used to lessen the amount of computing when the recommendation list need torecalculate because of the update of the user-item matrix (evaluating matrix). Thisalgorithm uses the idea of ‘a space for time’, simplifying the computation by cachingintermediate data, then we can enhance the system’s performance to some degree. Theexpriments on MovieLens’ dataset prove that the improved algorithm has betterprediction performance and efficiency than the traditional one.Cloud computing is one of the hottest buzzword in today’s IT industry. It’sconsidered as the core technology which will lead the next revolution of this industry.As the cloud has a very powerful calculation and storage capabilities, this dissertationponders how to use the advantages of cloud computing to solve the serious scalabilityissues which traditional collaborative filtering recommendation algorithm face, for thispurpose I adopt Hadoop, which is an open-source project of Apache Fundation, as the cloud development platform for my algorithm. The dissertation also do a little morethorough research on its distributed file system HDFS and MapReduce paradigm for thepurpose of implementing the above improved algorithm on this platform. Theexperimental result shows that as the machine number in the cluster increasing, theefficiency of algorithm also improving. This explains that the algorithm has an idealparallel performance and highlights the advantage of the combination of cloud platformand recommendation algorithm.