A Search Results-based Personalized Recommendation System

With the incessant development of computer technology and network technology, the information on Internet is increasing quickly. Facing large numbers of information, it is already more and more difficult for users to try to find and search information through scanning Web. They often spend much time, but gain little. Here search engines can be utilized to help to search useful information. The existing search engines, such as Google, Yahoo, Infoseek, often return a long list of search results. So the users have to validate those snippets of search results one by one to find whether they are the ones the users want, which is a time-consuming course.One method of solving the before-mentioned problem is clustering the search results through applying technology of Web mining; in this way the users can descry the results by groups. However, as non-state character of HTTP protocol, search engines can't track user's favor perfectly. Although some search engines can mark the search results and display it to users by rank, they commonly don't consider user's interests. Different users gain the same results with the same query keys. Further more, traditional clustering methods can't solve the problem of supplying search results by user's interests commonly.It is a remarkable problem that the amount of results returned by search engines is too large and the engines can't provide the users with required results based on their interest. An improved STC algorithm is put forward in this paper by combining user-interest model with STC clustering algorithm. In addition, a policy of personalization recommendation and a method of updating user-interest profile are also proposed at the same time. A personalization recommendation system based on search result (SRPRS) has been implemented by the use of the improved STC algorithm. SRPRS can organize search results automatically, and help the users find the needed results with the specific topics. Finally, the paper analyzes the clustering and time properties of the SRPRS by experiments. In allusion to the search result list, SRPRS has a better performance in fast finding the user's interested documents.2...