Materialized View Selection Using Memetic Algorithm

Abstract

A data warehouse stores historical data for the purpose of answering strategic and decision making queries. Such queries are usually exploratory and complex in nature and have high response time when processed against a continuously growing data warehouse. These response times can be reduced by materializing views in a data warehouse. These views, which contain pre-computed and summarized information, aim to provide answers to decision making queries in an efficient manner. All views cannot be materialized due to space constraints. Also, optimal view selection is shown to be an NP-Complete problem. Alternatively, several view selection algorithms exist, most of these being empirical or based on heuristics like greedy, evolutionary etc. In this paper, a memetic view selection algorithm, that selects the Top-T views from a multi-dimensional lattice, is proposed. This algorithm incorporates the local search improvement heuristic, i.e. Iterative Improvement, into the evolutionary manner for selecting an optimal set of views, from amongst all possible views, in a multidimensional lattice. The purpose is to efficiently select good quality views. This algorithm, in comparison to the better known greedy view selection algorithm, is able to efficiently select better quality views for higher dimensional data sets.