A Dynamic Adjustment Strategy Of Materialized Views Based On Query Clustering

With the coming of the information era,data warehouses are used widely.However,the operation of online analytical processing for a data warehouse is time-consuming.Therefore,the technology of materialized view was invented.It store the probable query results physically to improve the query response performance of the data warehouse.Nevertheless,it is unrealistic to materialize all views in a limited storage space.Selecting the valuable views to be materialized to improve the query response performance of the data warehouse in a limited storge space is mainly studied in this research.The followings are objectives of this research:A large number of user query records are accumulated during the operation of a data warehouse.These records reflect the user's attention to the data intuitively.A method of calculating the attribute similarity can be proposed based on algorithm of mining association rules.Then a method of calculating the query similarity can be proposed based on attribute similarity.Finally,candidate views set can be generated according to clustering the queries set.When there are too many candidate views,only the part of them can be materialized within the limited storage space.What's more,the value of the partial views are small because of inherent sparsity of multidimensional data set.So a cost model has been proposed applicable for the views with equal size.Combining the size of candidate views with the query frequency of the subview of the candidate view,the benefit of each view in the candidate view set can be calculated,then the candidate views with greater benefit can be materialized.Considering the characteristics of time-variant of Data warehouse.A dynamic adjustment strategy of materialized views based on query clustering(DSMVQC)is presented,user's query records can be collected regularly,then cluster the similar recoreds to generate candidate view set.The benefit of candidate views can be calculated by cost model.Under limited store space,select valuable views following some elimination rules dynamically.The experimental results based on the Airline Ticket Settlement Data Sets show that,DSMVQC strategy keeping the data warehouse with a high query response performance at different stages.