Experimental Reproducibility Package for DBSCAN-Assisted K-Means Energy Data Clustering
Year: 2026
Type: software
Abstract: This record provides the updated experimental reproducibility package for the study “A Hybrid Partition–Density Clustering Approach for Cluster Quality Enhancement under Uncertainty.” The package contains the complete Jupyter notebook, Python implementation, experimental configuration, generated result tables, cluster profiles, validation outputs, and publication-oriented figures used in the study. The experimental evaluation compares K-Means, Fuzzy C-Means (FCM), Trimmed K-Means, DBSCAN, and the proposed DBSCAN-assisted K-Means approach on four public energy-related datasets and one controlled synthetic dataset. The real datasets are obtained automatically from pinned public Kaggle versions, while the synthetic dataset is generated deterministically with known cluster labels and injected low-density noise. The complete Appliances Energy Prediction dataset is used without row-level subsampling. Clustering quality is assessed using the Silhouette Score, Davies–Bouldin Index, and Calinski–Harabasz Index. Additional analyses include retained-data coverage, full-data post-reassignment validation, Adjusted Rand Index (ARI) and Normalized Mutual Information (NMI) on the synthetic dataset, noise-detection precision/recall/F1, and normalized and original-scale cluster profiles. This updated version also includes a Trimmed K-Means robust baseline and full-data validation after reassignment of observations removed during density filtering, enabling a more balanced analysis of the trade-off between cluster quality and data coverage. The archive is organized into code, results, and figures directories and includes a README, requirements file, citation metadata, and file manifest. Raw public datasets and manuscript source files are not redistributed; dataset source links and version information are provided for reproducibility.
Institution: Azerbaijan Technical University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7214626432