LS-BMO-HDBSCAN as a hybrid memetic bacterial intelligence framework for efficient data clustering

Abstract

Data mining provides insights from massive and complex data to help make informed decisions, find hidden patterns, and solve difficult real-world problems. Unsupervised clustering groups natural data without labeling. Classic clustering algorithms like K-Means are popular due to their simplicity and computational efficiency, but they are sensitive to initialization and cannot handle noisy or non-convex cluster topologies. A hybrid clustering technique integrating L-SHADE, Bacterial Memetic Optimization (BMO), and K-means initialized HDBSCAN overcomes these concerns. The proposed system uses L-SHADE’s adaptive parameter management and convergence efficiency, BMO’s exploration–exploitation balance and memetic learning, HDBSCAN’s density-aware, noise-resilient clustering, and K-Means for centroid initialization Improve global and local search performance, convergence, avoid premature stagnation, and clustering in noisy, high-dimensional data with hybrid. Eleven popular benchmark datasets were utilized to evaluate LS-BMO-HDBSCAN. The recommended technique was compared against K-Means, PSO, NM-PSO, K-PSO, K-NM-PSO, CPSO, BFO, IBFO, BCO, and SMBCO. Performance was measured using Silhouette Score, Davies-Bouldin Index, Rand Index, Jaccard Index, and objective function. The durability, adaptability, and clustering accuracy of LS-BMO-HDBSCAN are confirmed by experimental results that outperform other methods across all datasets. This approach solves complex clustering problems in real-world data mining intelligently and reliably.