Research Of Collaborative Filtering Algorithm Based On Tag And Rating

In pace with the Internet and the development of Web2.0 technologies, the Intelligent Information Era is arriving with the mainstream feature of Big Data, Mobile Internet and Cloud Computing. This will cause the network resources grow exponentially and result in information overload. According to Internet is widespread and popular, Personal Recommendation can effectively solve the problem of information overload, which provides targeted services based on user profiles. Collaborative filtering recommendation technology just relies on score, and it can adapt to a variety of specific applications. However, recommendation accuracy is often affected by cold start, sparsity dataset, and low similarity calculation precision. For purpose of solving the above-mentioned problems, most of traditional collaborative filtering algorithms are fond of resolving them by user, item, and score. Nevertheless, recommendation algorithms based on tag are often solving it via user, item, and tag. But seriously, the current researches seldom make good use of tag which contains a large amount of user data such as interests, behavior, etc. The tag can be calculated by predicting the unknown ratings data. Therefore, in condition of poor sparse datasets and short of scalability, it is significant and crucial to research collaborative filtering algorithms based on tag and score.The main task is as follows:First of all, it summarizes the related knowledge of recommendation system, and describes the research status of scholars on collaborative filtering algorithm. By means of analysis collaborative filtering algorithms and it's existed serious problems. From the perspective of tag and score, this paper analyzes collaborative filtering recommendation algorithms in-depth.In the second place, taking into account diverse user profiles and target item can affect recommendation accuracy; we propose a collaborative filtering algorithm of selecting neighbors based on target item and user profiles.Ultimately, in certain sparse dataset circumstances, use tag cloud for reference, and tags are transform into score, then filling user item matrix. This is a collaborative filtering algorithm based on tag filling score. Experimental results show that the algorithm can effectively alleviate the sparsity dataset, while increasing the accuracy of recommendation.