| In the era of information, the amount of information that a single person can get is huge, which makes a great contribution to the technical progress of the society. However, it brings some inconvenience at the same time:how to find the very information that is valuable to us? Previous solution is developing search engines to search the information, but this scheme can do nothing to the hiding information with potential value. In order to remedy the defect, intelligent recommendation engine emerges as the times require.With the development of the data mining technology, more and more internet corporations start to offer users the applications of recommending items. When people are viewing the electronic commerce website, they are going to find some items marked "you may also interest in" in striking place of a page. When viewing the websites like imdb douban, people are going to find some movies marked "you may also want to watch". This is the intelligent recommendation function of the website.If items recommended to the users fit their preferences, causing their interests, users are very likely to find the right items or movies. The whole user experience of the website obtains a very big promotion, and the intelligent recommendation system plays a good role of leading.The basic idea of realizing this kind of recommendation system is building users’ preference model according to users’ existing interests and demands, thus comes to the items users may interested in. Comparing to the common searching engine, recommendation system offers personalized service, saving some trouble from users of searching items. It has very broad development and research prospects.What is realized in this paper is the most widely used collaborative filtering algorithm. The so-called’collaborative filtering’ is just like people would ask opinions from their friends and families when they are going to watch movies. That is finding some users similar to the target user in interest among the whole user group, or finding some similar items among the whole item group, then making recommendations with the help of them.The recommendation system realized in this paper uses Java for programming, and development environment is Eclipse integrated with open source projects Mahout and Maven. After realizing the CF recommendation system, this paper evaluates and compares existing CF recommendation algorithms’ performance. Then this paper designs a new algorithm I-ST-CF on the basis of existing algorithms combined with the recommendation based on content, which improves the cold boot problem and data sparsity problem. This paper verifies that the new algorithm can improve the performance of the recommendation by experiment. At last, in order to meet the demand of processing large data set, this paper realizes distributed calculation of the recommendation system based on Linux and Hadoop, and verifies that increasing the number of computers in the cluster can accelerate the calculation speed and enhance the efficiency.recommendation system, collaborative filtering, Mahout, Hadoop, distributed calculation... |
PDF Full Text Request