This repository contains the experimental data associated with the MCSA (Montecarlo-Clustering Search Algorithm), a stochastic metaheuristic designed for solving optimization problems. The dataset supports the results presented in the associated publication and includes raw outputs, benchmark evaluations, and problem-specific instances such as the Knapsack Problem (KP) and Multi-Objective Knapsack Problem (MOKP). Detailed tabular and file-specific descriptions are provided within each experiment folder (EX01_instance–EX04_Synthetic). For a deep dive into the algorithmic steps, binary encoding, and problem decomposition, please refer to the README_methodology

METHODOLOGICAL OVERVIEW

The dataset was generated using the MCSA (Montecarlo-Clustering Search Algorithm), which combines:

Montecarlo sampling for global exploration Clustering strategies for solution space refinement

Due to its stochastic nature:

Multiple independent executions are required Different random seeds produce different solution trajectories Performance must be evaluated statistically across runs

DATA TYPES INCLUDED

Plain text files (.txt): raw outputs from algorithm executions

CSV files (.csv): structured problem instances and reference solutions, and aggregated experimental data (e.g., MOKP and benchmarks)

Image files (.png): visualizations of results

REPRODUCIBILITY

Each execution in the dataset corresponds to an independent run of the algorithm.

File naming conventions include parameter settings and random seeds.

Raw outputs are provided to allow independent analysis.

Note: Source code is not included due to access restrictions. However, the dataset contains sufficient information to analyze the behavior and results of the algorithm.

USAGE NOTES

Users should begin by exploring the EXPERIMENTS directory (EX01_instance–EX04_Synthetic) for structured and well-documented cases.

The PARALLEL BENCHMARKS directory contains large-scale experimental data and may require additional processing.

The METHODOLOGY folder is intended for interpretability and documentation of the algorithm’s internal behavior.

MORE INFORMATION

Due to the stochastic nature of MCSA, not all executions converge to the optimal solution. Therefore, analysis should consider multiple runs and aggregated performance metrics rather than single execution results. The best solution is obtained by comparing the results across multiple independent executions and selecting the best-performing outcome.

For detailed descriptions of individual experiments, refer to the README files included in each experiment folder.

For a deep dive into the algorithmic steps, binary encoding, and problem decomposition, please refer to the README_methodology