Research On Materialized View Selection Algorithm In Data Warehouse

Data warehouse is subject-oriented, integrated, relatively stable and reflecting history change data set, contains many data that come from multiple distributed, autonomous, and possibly heterogeneous data sources. Data warehouse faces analytical data processing, and provides data supporting management decisions for online analytical processing (OLAP) and data mining. In order to enhance query response efficiency of data warehouse, store some intermediate results of searching in data warehouse, and this technology is materialized view technology. In the real applications, it is needed to determine which view to select according to query cost, storage cost and view maintenance cost, which is materialized view selection problem this paper needs to solve.In order to solve materialized view selection problem, this paper uses multi-dimensional data lattice to organize view, and determines cost model of maintaining—query according to actual situation of computer development, that is to select materialized view set which makes maintaining cost minimum when query cost is less than given constant. And materialized view selection problem studied by this paper is called MQ_MVS for short.This paper inauguratedly introduces clonal selection algorithm to solve MQ_MVS and presents CSA_VSP algorithm. CSA_VSP algorithm eventually obtains materialized view set meeting cost model through interactive process of clone, variation and selection operators. This article compares CSA_VSP and genetic algorithm (GA) through test, and makes a conclusion that solution quality of CSA_VSP is superior to that of GA on solving MQ_MVS. But in the variation process of CSA_VSP, this algorithm can only introduce new antibody, and some individual antibody may appear degradation. During the process of CSA_VSP variation, this paper introduces Metropolis criterion of simulated annealing algorithm that could improve running efficiency of algorithm on the premise of ensuring the diversity of antibody, and presents improved CSA_VSP algorithm namely clonal selection and simulated annealing algorithm (CSSA_VSP). This article compares CSSA_VSP and CSA_VSP through test, and makes a conclusion that solution quality of CSSA_VSP is superior to that of CSA_VSP on solving MQ_MVS.