A Practical Methodology for Reproducible Experimentation: An Application to the Double-Row Facility Layout Problem.

Area of Science:

• Computer Science
• Operations Research
• Artificial Intelligence

Background:

• Reproducibility is a significant challenge in stochastic methods like evolutionary algorithms and metaheuristics.
• Existing literature offers general guidelines but lacks practical steps and software tools for enhancing reproducibility.
• Stochastic optimization methods are widely used across various scientific and engineering domains.

Purpose of the Study:

• To propose a practical, step-by-step methodology to improve the reproducibility of experiments using stochastic optimization methods.
• To develop and integrate software tools that support the proposed methodology, implementing state-of-the-art techniques.
• To address the specific challenge of reproducibility in the context of the double-row facility layout problem (DRFLP).

Main Methods:

• A three-step practical methodology designed to enhance experimental reproducibility.
• Development of accompanying software tools that guide researchers through the methodology.
• Application of the methodology to the double-row facility layout problem (DRFLP), including replication of prior methods.

Main Results:

• A novel algorithm for the DRFLP that outperforms existing state-of-the-art methods.
• Successful application of the proposed methodology to a complex optimization problem.
• Replication of previous methods on larger instances to provide a comprehensive comparison.

Conclusions:

• The proposed methodology and associated software tools significantly improve the reproducibility of research involving stochastic optimization.
• The developed DRFLP algorithm offers superior performance compared to existing approaches.
• All artifacts, including the methodology and study data, are publicly available on Zenodo to foster further research and verification.