Empirical analysis of metaheuristic search techniques for the parameterized dynamic slope scaling procedure

TL;DR

This paper empirically evaluates metaheuristic algorithms like simulated annealing, tabu search, and particle swarm optimization to optimize the parameter in the dynamic slope scaling procedure, enhancing solutions for the fixed charge network flow problem.

Contribution

It introduces a metaheuristic-based approach to optimize the parameterized dynamic slope scaling procedure, improving solution quality for the FCNF problem.

Findings

Metaheuristics effectively improve solution quality.

Solution improvements are robust across different metaheuristics.

The approach handles complex FCNF problems efficiently.

Abstract

The dynamic slope scaling procedure is an approximation method successfully which solves the fixed charge network flow (FCNF) problem by iteratively linearizing the fixed cost. The parameterized dynamic slope scaling procedure adds an additional $\psi$ parameter to the procedure which can significantly improve the solution quality. Finding the optimal value of $\psi$ for a given problem is non-trivial. This paper employs multiple metaheuristic techniques, including simulated annealing, tabu search, and particle swarm optimization, to guide the search for good parameter values. In rigorous testing, we examine the search results, compare the improvement efficiencies among the techniques, and evaluate the final solution quality of the FCNF problem. The experiments show that the solution improvement is robust with respect to these metaheuristics and the complexity of FCNF problem.