Genetic algorithm for quadratic assignment problems: application of Taguchi method for optimisation. (31st March 2020)

• Record Type: Journal Article
• Authors: Pradeepmon, T.G.; Panicker, Vinay V.; Sridharan, R.
• Is Part Of: International journal of operational research. Volume 38:Number 2(2020)
• Page Start: 193
• Page End: 220
• Publication Date: 2020-03-31
• Subjects: quadratic assignment problem -- QAPs -- genetic algorithms -- Taguchi's design of experiments method -- optimisation of operations and parameters; Operations research -- Periodicals; 003.05
• Languages: English
• ISSNs: 1745-7645

Abstract

Quadratic assignment problems (QAPs) are the hardest of combinatorial optimisation problems, with some problems of sizes of the order of 30 still remaining unsolved optimally. Solving QAPs with exact optimisation methods is cumbersome and hence, the use of non-conventional optimisation methods is recommended. Genetic algorithm (GA) being one of the most popular evolutionary algorithms is an appropriate choice for solving QAPs. The methods of operations used in GA influence the solution quality and thus, an optimal combination of parameters and operators are required for the efficient implementation of the algorithm. In this paper, the Taguchi's design of experiments method is used to find the best parameter combination and the best performing combination of operations for GA. The GA thus obtained by incorporating the selected parameter values and operators is then used for solving the QAPs taken from the QAP library. For many of the problems, it is found that the results obtained are within one percentage deviation from the best-known solutions.