An innovative approach, k-means with foggy centroid selection (KFCS) was proposed, for enhancing data clustering performance. This study focuses on the application of this method to the Pima Indians diabetes database, serving as a comprehensive evaluation ground. The process begins with preprocessing and data arrangement, involving scaling and normalization to ensure accurate computation. KFCS, combines k-means clustering with foggy centroid selection, utilizing both random initialization and iterative centroid calculation. The approach hinges on four distance algorithms – Euclidean, Pearson Coefficient, Chebyshev, and Canberra – to gauge similarity. A detailed exploration of distance estimation enhances dataset understanding. Through rigorous evaluation, KFCS demonstrates superiority in terms of computation time and error analysis, with Canberra algorithm emerging as a standout performer. This work contributes a comprehensive methodology for improved data clustering and analysis.