A Study Of Search Ranking Algorithm Based On User Behavior Analysis

Currently the major search engines primarily through mass action to determine theimportance of a web page, although some of them adopted some factors which based onpage views, citations, PageRank algorithms such as weights to improve search results.However, it still unable to meet the individual’s personalized search. Searched contentcan not be sorted according to personal interests or personal knowledge background.In order to solve the above problems which can not provide for the individual needs.Search ranking algorithm based on user behavior proposed here. It will construct userbehavior model by the algorithm, using the user behavior model to calculate the user ’sdegree of interest on the page. Using the degree of interest to page and the weight searchengine returned to a comprehensive re-sort. Experiments show the average case to begood that adding the user’s own conduct search.Sorting the search results based on the user behavior mainly centered in the userbehavior are better in most cases. The effect of getting are better especially for those whohave some experience in search or a relatively high search level users.. But in the actualprocess, we found that the effect of the new users whom not added to user behavior areworse. And some getting worse. To solve this problem, this references expertscharacteristic. Users will be clustered with the same interest. And then extracted expertsfrom category. With these experts who have a high level of skill to help the averagesearch user. The original NOREN clustering algorithm was improved to obtain expert.This algorithm is validated through experiments, the average search efficiency thanindividual user behavior only added to the user’s own algorithms to be high.To validate the algorithm. We have developed a prototype system.To carry outexperimental Through behavioral simulation users and experts，the results show that asearch engine user adding behavior than did not has been significantly improved. Whilethe average search efficiency with expert than only individual behavior to be high.