A game theory based framework for materialized view selection in data warehouses. (May 2018)

• Record Type: Journal Article
• Authors: Azgomi, Hossein; Sohrabi, Mohammad Karim
• Is Part Of: Engineering applications of artificial intelligence. Volume 71(2017:Nov.)
• Page Start: 125
• Page End: 137
• Publication Date: 2018-05
• Subjects: Materialized view selection -- Multi-value processing plan -- Data warehouse -- On-line analytical processing -- Game theory; Engineering -- Data processing -- Periodicals; Artificial intelligence -- Periodicals; Expert systems (Computer science) -- Periodicals; Ingénierie -- Informatique -- Périodiques; Intelligence artificielle -- Périodiques; Systèmes experts (Informatique) -- Périodiques; Artificial intelligence; Engineering -- Data processing; Expert systems (Computer science); Periodicals; 620.00285
• DOI: 10.1016/j.engappai.2018.02.018
• Languages: English
• ISSNs: 0952-1976

Abstract

Abstract: Data warehouses exploit On-Line Analytical Processing (OLAP) to make rapid answers for analytical queries. Huge amount of aggregated data within a data warehouse on the one hand, and complex analytical queries raised in a data warehouse on the other hand, increase response time to queries extremely. To solve this problem, a number of views are derived and extracted from original base tables and queries have been answered using them. Since materialization of all possible views is not effective because of limitation of storage and maintenance overhead, selecting an optimal set of views for materialization is crucial to maximize data warehouse performance. In this paper, a game theory based framework for the materialized view selection is proposed. In the proposed framework, query processing and view maintenance costs play a game against each other as two players and continue the game until reach the equilibrium. According to the framework, a new static method, called Game Theory based Materialized View selection (GTMV), has been proposed. Verification of proposed approach has been evaluated using several synthetic and real world datasets. Experimental results show that the GTMV method has better performance comparing previous algorithms and substantially outperform former methods.