Handling Constraints Using Penalty Functions in Materialized View Selection

Abstract

Materialized view selection (MVS) plays a vital role for efficiently making decisions in a data warehouse. This problem is NP-hard and constrained optimization problem. The authors have handled both the space and maintenance cost constraint using penalty functions. Three penalty function methods i.e. static, dynamic and adaptive penalty functions have been used for handling constraints and Backtracking Search Optimization algorithm (BSA) has been used for optimizing the total query processing cost. Experiments were conducted comparing the static, dynamic and adaptive penalty functions on varying the space constraint. The adaptive penalty function method yields the best results in terms of minimum query processing cost and achieves the optimality, scalability and feasibility of the problem on varying the lattice dimensions and on increasing the number of user queries. The authors proposed work has been compared with other evolutionary algorithms i.e. PSO and genetic algorithm and yields better results in terms of minimum total query processing cost of the materialized views.