Optimizing Local Search Ranking Algorithm By Using User Behaviors

The search engines have only twenty years’ history, but have developed from theserver search to the online mass data search. Now, thanks to gradually improverequest of the quality of search results and the rapid increase in the amount of networkinformation, the search engine started develop toward specialization andhumanization.Local search engine as a general search engine personalization to achieve, it isvery different in the retrieval of content and data structure from general search engine.So traditional ranking algorithm cannot be fully applicable to the local search. Thecontent of the local search is closely related to people’s daily lives and the retrievalprocess is largely dependent on the user’s behavior. According to the features of thelocal search, through comparing and analyzing the user search behavior in generalsearch and local search, this dissertation presents a local search engine systemframework that is based on features of user behavior.First, the Nutch search engine is used as part of the core foundation. Thedictionary of the local community is added, the two-way matching method is used forChinese word segmentation, the POI three-part index is proposed for local search,which implement the three basic modules of the local search engine: crawlers,indexing and retrieval module.Then, analyze research development situation of the current general searchusing user behavior features optimization search and analyze the method ofacquisition in existing user behavior features. After comparing the advantages anddisadvantages of local search and general search, select user behavior features whichapply to local search, and put forward the local search user behavior featuresacquisition scheme.Finally, in order to use the extracting local search user behavior features toranking optimization, change the SVM classification algorithm to a rank SVMalgorithm which is suitable for ranking. And then using SVM ranking algorithm fusesthe eigenvalues of local search user behavior and the ranking results of general search,it is the local search ranking results based on user behavior analysis.Three methods of contrast experiment is carried out to verify the role the userbehavior plays in the optimization of the rank results. Experimental results show that after integrated into the user behavior features, the average accuracy rate of the localsearch rank results and the top ten documents dependency have improved to someextent. User behavior features allow ranking results can more easily and accuratelyresponse the user’s interesting, improve the user’s search experience.