Research On Query Optimized Techniques In Spatial Data Warehouse

The performance of the query restrained the using of spatial data warehouse severely. Based on the background of the real application in the spatial data warehouse of Xuzhou branch of China Construction Bank, this paper analyses the features of spatial data and SOLAP query in spatial data warehouse, focuses on the problems of its query function, and studies the optimization of the Materialized View Query.This paper analyzes the spatial range aggregate queries, and presents the two-stage materialized method. It processes efficiently range aggregate queries on spatial and non-spatial dimension. The method main idea is as follows, to form the non-spatial Range Aggregate Queries which are frequently used into a candidate view collection without considering the spatial queries, use genetic algorithms to select the view which costs minimal, and materialize the view. This process is called Primary Materialize. After then, calculating the value of each intermediate node in the R-tree in all the materialized view, and saving all the values in the correspondent table. This process is called two-stage Materialize. The Secondary Materialized View contains the results of the intermediate node in the R-tree, so the access times of spatial queries diminished, the query time of primary view shortened, and the efficiency of the aggregate query improved accordingly.This paper raises a Similarity Measure Function used for multi-dimension, sparse, and binary data based on the intensive analysis for the aggregate techniques; bring forwards spatial Aggregate Greedy Algorithm aimed at the maximal indirect profit in spatial greedy algorithms. The algorithm finds out the spatial object group first in adjacent spaces, that is combination group. The spatial objects in the group can be combined to one. The Algorithm calculates the profit of the combination group in every clustering, rather than calculate it in the whole group. Besides, find the combination group with the maximal profit in the clustering, save the group and its profit. After select the combination group, recalculation for the profit of the combination group in other clustering is not needed, so the calculating work is simplified largely. The effectiveness and advantages of this algorithm is proved by simulations.This paper presents the Dynamic View Selection Algorithm based on Cost Model of spatial data warehouse. The algorithm preserves some storage space to store new views. If more space is need, delete the less profitable views one by one when the preserved space is used out. Repeat the above step till the need for the idle space is fulfilled. Experiments proved the validity of the algorithm and compared the performance of the algorithms in different preserved time.