Clustering Algorithm Selection Benchmark: 295 Datasets, Meta-Features, and Internal Validation Metrics

Description This research dataset supports the study of automatic clustering algorithm selection using dataset meta-features, internal validation metrics, and genetic algorithms. The collection comprises 295 tabular datasets characterized through 36 meta-features describing statistical, structural, geometric, correlation, density, and principal component analysis properties. The repository includes dataset classifications by size and dimensionality, extracted meta-features, clustering evaluation results, and internal validation metrics, including Silhouette, Davies–Bouldin, Calinski–Harabasz, and Dunn. These resources support the construction of meta-datasets for supervised learning, where the target variable represents the best-evaluated clustering algorithm according to the corresponding internal validation criterion. The research considers two experimental strategies: a weighted combination of internal validation metrics and independent meta-datasets for each metric. Genetic algorithms are employed for meta-feature selection in the second strategy. The dataset is intended to support reproducibility, comparative analysis, and further research on meta-learning and automated clustering algorithm selection.