Research And Implementation Of Personalized Search Engine Based On Query Preference

Nowadays,the rapid development of Internet has brought convenience to people’s life.At the same time,it’s not easy to obtain the information needed to users.Therefore,the emergence of search engine has changed the way of finding information.As a common way of network information retrieval,it has become an important tool for every user who surf the Internet,it has been widely concerned and used by people.However, traditional search engine also exists many deficiencies.It is mainly because the traditional search engine adopts keyword retrieval method, the lack of use of the user’s personal information, which cannot be targeted to provide users with personalized service.So, based on the consideration and analysis of different information needs of users, personalized search system came into being, it appears to some extent to meet the user’s personalized information needs.In this paper, we used the information recommendation technology,which widely used in the electronic commerce website, and proposed a feasible scheme of personalized information service, that is to use the query recommendation technology in user’s search to realize the personalized search recommendation.Related research shows that the user’s query history reflects the user’s search habits and query preferences, so this paper made a thorough analysis of the user’s historical query data, and proposed a user click model to predict user queries and eventually made recommendations.The research in this paper is mainly focused on the query recommendation. Firstly, according to the user’s search click history data, a user click model is trained by using the Naive Bayes theory.this model is used to analyze the query of user and predict CTR(click-through rate) between the user input and links.Then the average CTR was assigned to the corresponding query items based on the bipartite graph,to sort the queries, the highest K queries are recommended to user.Secondly, on the basis of the historical data of a single user, this paper proposed the user query recommendation technology based on collaborative similar calculation, expanded the current user dataset based on the users which have the same search behavior. This method can solve the problem of insufficient data for the current user,and it can also provided users with a certain extension and novelty in query recommendation.In the similarity computation of the user,this paper took the log of each user as a document,and uses vector space model to calculate the similarity between the users’documents.The frequency ratio of user link-click is considered as the preference score of each link, and then the improved Euclidean distance is used to calculate the user’s nearest neighbors.Finally,the two methods are combined to calculate the similar users of the current user, and eventually generated recommendations based on the click model.Based on the research of the query recommendation, this paper realized a simple search engine system. The related recommendation algorithm is applied to the system, and the user click model is used to the web page ranking, the personalized query recommendation and web page ranking function is realized.