<![CDATA[This study develops Stability-Dispersion Adaptive Weighted K-Means (SDAW-K-Means), an extension of classical K-Means that updates feature weights according to within-cluster dispersion. Classical K-Means treats all standardized features equally, although some features may be more relevant for cluster separation than others. The proposed method estimates feature weights iteratively: features with smaller within-cluster dispersion receive larger weights, while less informative features receive smaller weights. The empirical illustration uses the public Iris dataset from the UCI Machine Learning Repository through scikit-learn. Results show that the proposed weighting mechanism is interpretable and can improve agreement with reference labels based on the adjusted Rand index. The article contributes a transparent feature-weighted K-Means formulation for applied clustering research.]]>
Copyrights © 2026
