Research On Recommendation Algorithm Based On Popularity And User Preference

With the rapid development of information technology, the problem of " information overload " caused by massive information has become more and more serious, which makes people lost in the ocean of information and it’s difficult to obtain the information they need. How to mining the valuable information quickly and effectively becomes the key to solve all kinds of problems. Recommendation system as a kind of mature personalized service technology has been widely used in e-commerce and other fields.The recommendation technology is the key of the recommendation system. The common recommendation technologies include:collaborative filtering recommendation, model-based recommendation, content-based recommendation and hybrid recommendation. Among them, the collaborative filtering algorithm is the most widely used. On the basis of summarizing recommendation technology, this paper focuses on research and improvement of data sparsity and cold-start problem in collaborative filtering algorithm. The content mainly includes:(l)This paper analyzes the research status and development of existing recommendation technology and summarizes the merits and defects of various kinds of recommendation technology. Then we introduce the theory of item-based collaborative filtering and user-based collaborative filtering algorithm. After that the concrete steps of realizing the algorithm are given focusing from item-based recommendation algorithm.(2)We introduce the related content of the cloud model and explain the cloud model based collaborative filtering algorithm in detail. In addition, the proposed algorithm can comprehensive the item popularity and the users’preference on item attribute, which come to whole preference degree on item. The algorithm filling the unrated-items with the sum of preference degree and average score of users, Meanwhile when meet with the problem of cold-start, we can use the users’preference degree on item to predict rating about the new items.(3)This article taking advantage of mapReduce parallel processing framework and HDFS in storage safety% stability, we design and realize the recommendation algorithm using mapReduce. Then with the real dataset we complete the recommendation of tourist spots under the cloud environment.(4)Through comparative experiment this paper verifies the performance of the proposed collaborative filtering algorithm. The experiment takes the two kinds of data sets:MovieLens data provided by the GroupLens and the data of travel sites crawled by the LocoySpider. From the experiment results, we can analysis and summarize that the proposed algorithm has improved significantly in terms of accuracy and efficiency compared with traditional collaborative filtering algorithm.