Research On Selection Algorithm Of Materialized View In Data Warehouse

A data warehouse is, by definition,a subject-oriented, integrated, time-variant collection of data to enable decision making across a disparate group of users.Decision supporting applications involve complex queries on data warehouses, which contains large amounts of data.How to efficiently manage so many data is one of the problems faced by data warehouse, and materialized view is the important method to solve it, but it need extra system space to store, and it need maintenance at price of system cost, so it is an important reaserch subject to choose which views to materialize.The thesis mainly works on the selection of materialized view, from two aspect, static and dynamic.The static selection algorithm of materialized view is based on the situation that the probability of queries has been provided by the user or the queries are distributed evenly.Firstly the thesis introduce three static selection algorithm of materialized view, and analyse their advantages and disadvantages; then a better cost model is put forward, it doesn't consider the query cost only, but also contain the update cost; at last at the base of the cost model, put Genetic Algortithm on Materialized View forward, it belongs to the static selection algorithm.In reality, the situation that the queries are distributed evenly is not the case, so the dynamic adjustment of materialized view is introduced.In this chapter, firstly it introduce the reaserch actuality about the dynamic adjustment of materialized view, and analyse their advantage and disadvantage; then because of the multiformity, we put a new algorithm forward, named Rough Set Clustering-Based Dynamic Materialized View Algorithm, and RSCMVDA is its short form, the algorithm is based on rough-set clustering on materialized view, execute the dynamic adjustment, it doesn't only satisfy the multiformity of the queries, but also consider the hiberarchy of dimesion; At last from the experiment we can get the conclusion, that as the queries set is increasing, the combination between GAMV and RSCMVDA is better than the genetic algorithm, that is because as the queries set is increasing, the probability of queries changed easily.