Research Of Materialized View Selection Algorithm Of Multi-dimensional Data In Data Warehouse

Data warehouse(DW) and On-Line Analytical Processing (OLAP) is an important way to gain decision-making support for corporation. Because it often access millions of data, and involve joins and aggregations operation when analyzing multi-dimensional data for various dimension. So how to reduce the response time for queries and improve the query efficiency become very important. Materialized views use some intermediate results from query processing stored in the DW to improve query execution speed, but it need to take up storage space and also bring system cost for view maintenance. Therefore, how to select a group of materialized views under the limit space to maximize improve the system response performance for queries is the focus of our research. Usually, we first to materialize part of views using the user's estimate query, and then dynamically adjust them according to the queries when system running.The paper first make improvement in static selection process of materialized views, proposing a view selection algorithm based on ant-colony and genetic algorithm. During ant colony evolution, the best path and the worst path is more important. it take the largest and the smallest pheromone update methods to point out the ant evolutionary direction which can effectively prevent blind search and improve problem-solving speed. In addition, it import cross operator and mutation operator after every evolution which broads the global search ability of the algorithm, both the accuracy and speed are greatly improved. The test results show it efficiently improved solution convergence rate and successfully resolved the problem that ant colony algorithm falls into local optimal solution easily.In the dynamic adjustment process of materialized views, the paper propose an improved strategy for dynamic batch adjustment of materialized views which is not make adjustment after a query is executed, but first collects queries in a statistical time, regarding the whole adjusts. The algorithm judges the query set first whether to satisfy the adjustment condition, if satisfies, product candidate view according to view visit frequency, and then use view batch selection algorithm, if not satisfies, then use view dynamic adjustment algorithm. The algorithm does not need to calculate frequently, and can keep a high response to user's queries due to the query set connect to the trend of user's query. The experimental result indicated that the algorithm can effectively avoid"the vibration"and enhance the stability of materialized views collection.