The determination of the most suitable inertia weight strategy for particle swarm optimization via the minimax mixed-integer linear programming model

Abstract

Purpose

This paper aims to consider each strategy of the particle swarm optimization (PSO) as a unit in data envelopment analysis (DEA) and uses the minimax mixed-integer linear programming DEA approach to find the most suitable inertia weight strategy. A total of 15 inertia weight strategies were empirically examined in a suite of 42 benchmark problems in the view of DEA.

Design/methodology/approach

PSO is very sensitive to inertia weight strategies, and therefore, an important amount of research attempts has been concentrated on these strategies. There is no research into the determination of the most suitable inertia weight strategy; however, there are a large number of comparisons related to the inertia weight strategies. DEA is one of the performance evaluation methods, and its models classify the set of strategies into two distinct sets as efficient and inefficient. However, only one of the strategies should be used in the PSO algorithm. Some effective models were proposed to find the most efficient strategy.

Findings

The experimental studies demonstrate that an approach is a useful tool in the determination of the most suitable strategy. Besides, if the author encounters a new complex problem whose properties are known, it will help the author to choose the best strategy.

Practical implications

A heavy oil thermal cracking three lumps model for the simplification of the reaction system was used because it is an important complicated chemical process. In addition, the soil water retention curve (SWRC) plays an important role in diverse facets of agricultural engineering. As the SWRC can be regarded as a nonlinear function between the water content and the soil water potential, Van Genuchten model is proposed to describe this function. To determinate these model parameters, an optimization problem is formulated, which minimizes the difference between the measured and modeled data.

Originality/value

In this paper, the PSO algorithm is integrated with minimax mixed-integer linear programming to find the most suitable inertia weight strategy. In this way, the best strategy could be chosen for a new more complex problem.

Citation

Özsoy, V.S. (2021), "The determination of the most suitable inertia weight strategy for particle swarm optimization via the minimax mixed-integer linear programming model", Engineering Computations, Vol. 38 No. 4, pp. 1933-1954. https://doi.org/10.1108/EC-05-2020-0272