Dynamic Materialized View Selection Research Based On Clustering

Through many years of development,data warehouse has been widely used in various industries.As the passage of time,the amount of data grows rapidly.In order to resolve the problem of increasing response time required by query,materialized view technology came into being, and has become a research hotspot in data warehouse.The materialized view technology stores the corresponding data physically,and speeds up the query response time through precomputation.However,its own need to spend a lot of resources.So how to choose a group of appropriate views to be materialized becomes an important issue which solves the problem of data warehouse query.Existing technologies of materialized view selection are most of the static algorithms,and conflicts with dynamic characteristics of decision support system.Nevertheless the dynamic algorithm of materialized view selection has less researches and shortcomings of excessive spending.Based on their shortcomings and former researches,the paper presents and implements one dynamic materialized view selection based on clustering.The algorithm uses static materialized view improved algorithm and clustering improved algorithm that are presented.The paper researches dynamic materialized view selection based on clustering and presents a relevant algorithm-CBD-MVS based on discussing data warehouse,materialized view selection an clustering analysis technology and so on.The algorithm uses clustering technology for clustering user query sentence in data warehouse,then merges user query sentence inside clustering after clustering and gets less candidate materialized views,finally selects a suitable static materialized view selection algorithm in order to get final materialized views.The main research contents is listed here:1.In order to overcome the shortcomings of the existing clustering algorithms that uses clustering to deal with the user query,the frequent closed itemsets is applied to cluster analysis technology.Through the implementation of frequent closed itemsets mining algorithm on the user query,the association rules based on the attribute fields are obtained.And by these association rules,we can get correlation matrixs and eigenvectors of the attribute fields,calculate the similarity of the attribute filed sets,perform k-means clustering algorithm to obtain the results of clustering.Experiments show that the method can achieve better result.2.Discussing data warehouse technology and materialized view technology and researching static materialized view selection algorithm-Greedy, BPUS and PBS and analysising their lack,then presenting one improved algorithm-BGA.The algorithm uses heuristic searching algorithm for searching pane chart,existing dependences among data cube chart,and makes use of cost model to filter materialized views witch have most benefit,then regards memory space and increased benefit as threshold,and gets same effects of query cost with BPUS,whereas time what are used is less obviously.Experimental results demonstrate its efficiency.3.Study dynamic materialized view selection in data warehouse and present one dynamic materialized selection algorithm based on clustering for the lack of existing materialized view selection.The algorithm uses clustering algorithm based on frequent closed itemsets to cluster user query,and view merging algorithm to establish candidate materialized views,finally makes use of imprved static selection algorithm-BGA to get materialized views.Experimental results demonstrate its efficiency and viability.