Based On Historical Query Relational Database Query Optimization Keyword Research Questions

Database (database, referred to as DB) has been widely applied to people’sproduction and life, it can efficiently support storage and query of structured data, butit requires the user to understand the underlying database model knowledge and masterStructured Query Language, which is a difficult thing for inexperienced users. On theother hand, the Internet information retrieval technology only requires the user to inputkeywords, the page will return relevant results containing these keywords to the user.This search technique queries unstructured data in terms of keywords, the result isoften imprecise and incomplete. The Structured Query support for efficient retrieval ofstructured data, and have the perfect query optimization techniques. Therefore, basedon relational database keyword search field emerged.Since2002, the relational database keyword search problem has graduallybecome a hot topic in the field of information retrieval, with the combination ofefficient database search and simple operating character of information retrievaltechnology, therefore, a relational database keyword search technology has beenwelcomed by users. This paper study a relational database keyword search problem,based on analyzing and exploiting the history of previous user queries, so improve thecurrent query result with using historical information. In this paper, the research resultsand contributions as follow:First, for a query, we calculate the current query results by using of certainhistorical query results. By the definition given in the query similarity, to find the mostsimilar historical query which is with the current query, the results of the query toreconstruct the history of the formation meet the results of the current query. Theexperiments show that the keyword query efficiency of utilizing historical informationis better than querying directly from based database, it is especially better while a smaller amount of data.Second, it recommends the current query by utilizing historical results. It buildbipartite graph model and compute the similarity between query and historical queryby combining historical results and historical queries, it recommends the connection ofhistorical query results to the current user by the similarity. This is a kind of indistinctrecommendation, our recommendation algorithm satisfies user as much as possible.Just as in e-commerce recommendation system, products are recommended to thecurrent user according to past consumption records, Although the products whichrecommended may not be adopted by user’s, however, they must be which user aremost interested in and most likely to accept.Similarly, recommending to current queryby using the results of history query is also the result which user are most interested inand most likely to accept. After a series of experiments, in numerous recommendedresults, user adoption rates above90%.Several results of history query are connected to meet the current query results.Keywords in these history query contains the keywords in current query. The mainsolution is to split the keyword in current query into two or more existing historyquery.Two or more results of history query are connected by RC algorithm to form theresults of the current query. RC algorithm and reconstruction algorithm are comparedby changing different parameters, their advantages in different situations are concluded.At the same time, we compare the KWSBH which proposed in this paper with theexisting BANKS by changing the parameter settings (such as the number of keywords,data set size, etc.),finally,we get the conclusion which the efficiencyof KWSBH arehigher than that of BANKS.