Research On Techniques Of Real Estate Information Personalized Retrieval And Recommendation

With the coming of information age, Real Estate Industry information as a part ofnetwork information data manifests an explosive growth throughout Internet that impedes thespecific search to satisfy different users’ personal requirement, since the tradition SearchEngines always provide the same results for different users through the passive and ossifiedsearching process according to the same keyword, without the consideration of the individualpreference.For the information search issues stated above, this paper introduces a Real EstatePersonalized Information Retrieval and Recommendation based on User Preference Modeland implements the relevant system, through the exploratory research on the followingaspects.Firstly, based on the review of the interrelated literatures at home and aboard, this paperputs forward research model and hypothesis to address the issues of the Real Estate Industryinformation search, combing the Information Retrieval technology, User Modeling technologyand Recommendation technology.Secondly, the User Preference Model was studied extensively. This paper proposes toguide the update of the model by Fitness Function through learning the query and theoperative behavior of search results; meanwhile, considering the feature of Real EstateIndustry information, this paper adopts the Vector Space Model and designs a featuredictionary to achieve the mapping from query to user preference and get the specific weights,which facilitates the similarity computing algorithm to optimize the sorting of the searchresults.Thirdly, on the basis of the User Preference Model above and the advantages anddisadvantage analysis of Content-based Recommendation and Collaborative FilteringAlgorithm, this paper proposes an Adaptive Recommendation Algorithm combingContent-based Recommendation with Item-based Collaborative Filtering Recommendation,with the addition of a data structure—latest approved queue, which provides the basis for thecombination of these two algorithms and changes the parameter adaptively to promote thequality of recommendation.Finally, applying all the theoretical knowledge above, the Real Estate PersonalizedInformation Retrieval and Recommendation system based on User Preference Model is achieved. Through the experiment of retrieval performance test using P@N index, it provesthe application of User Preference Model improves the retrieval performance significantly,especially when the User Preference Model is initialized, as the P@N raises about25%,comparing the traditional Real Estate Personalized Information Retrieval. Meanwhile, basedon the former experiment, the recommendation performance is tested by usingPrecision-Recall index that proves the Adaptive Recommendation Algorithm combingContent-based Recommendation with Item-based Collaborative Filtering Recommendationperforms better than the original Collaborative Filtering Algorithm.