Materialized view selection using discrete genetic operators based particle swarm optimization

Abstract

The purpose of a data warehouse is to support decision making. Decision making queries posed against a large data warehouse, are typically analytic, intricate and recurring in nature. Such queries contain a lot of join operations over large volumes of records. Answering these queries efficiently is a critical concern in the data warehouse environment. One way to address this problem is by materializing views in the data warehouse. All views cannot be materialized due to storage space constraint. Materializing an optimal subset of views is shown to be an NP-Hard problem. Alternatively views can be selected either empirically or based on some heuristic like greedy, randomized, evolutionary, swarm based etc. In this paper, a discrete genetic operator based particle swarm optimization (DGPSO) has been used to select Top-K views from a multidimensional lattice. Further, experiments based comparison of the DGPSO based view selection algorithm with the fundamental view selection algorithm HRUA shows that the former is able to select better quality views for materialization.