Data Cube Investigation In The OLAP

With the development of database technology and the expansion of database scale, we hope to refine the useful information from existing data to serve for decision. Data warehouse technology and OLAP(Online Analytical Processing) are important new technology in dealing with business data in recent years, which have developed gradually in order to meet this technology. OLAP organizes multidimensional data by data cube model, which is convenient to queries. OLAP usually involves complex queries in large multidimensional datasets. Running analytical queries directly by the basis data volume of data warehouse will result in unacceptable quering performance. The solution to the problem is storing materialized view in the warehouse, which pre-aggregates the data in order to avoid raw data access and speed up queries.But there are many difficulties in pre-aggregating data cube (selection of materialized data cube view). How to select data cube to materliaze and which data cube will be selected are important problems. At present the study of selecting materialized data cube view developed few years in the academic circles, there are two important algorithms: greedy algorithm and genetic algorithm. On one side greedy algorithm only can get local solution, and hardly get globle solution; on the other side, genetic algorithm can not get better result in getting globle solution because of the characteristic of earlier converging. For this condition, this paper proposes a genetic algorithm which was improved. We use it to compute the selection of materilazed data cube.This paper focus on selecting materialized data cube, taking data cube part materialization as the object of study, and proposes an improved genetic algorithm to compute the materialized data cube when we build the data warehouse, we design a searching lattice model according to the reality needs. On the basis of this model we use the improved genetic algorithm to be in the progress of studying.We insert greedy algorithm into genetic algorithm's decoding, which improves its adaptional ability. This method can keep the multiplicity of population and also can speed up queries. We compare this algorithm to greedy algorithm and traditional genetic algorithm, the result is that the one is optimum.Finally, this paper points out the challenges of the selection of materialized data cube and summarizes the work in the future.