<![CDATA[Categorical student segmentation often suffers from bias due to subjective cluster (K) determination and vulnerability to local optima. This research proposes a robust comparative framework among K-Means, K-Modes, and K-Medoids to optimize university promotion strategies. Model robustness was tested via 30 independent iterations and evaluated fairly in an "Equivalent Evaluation Space" based on Hamming distance. Optimal architecture determination was automated using Multi-Criteria Decision Making (MCDM) integrating Borda Score and Z-Score Composite. Regarding computational Output, K-Means (K=10) dominated with the strongest cluster density (BetaCV=1.8705), highest initialization stability (ARI=0.5010), and maximum Borda score (12 points). As a managerial Outcome to mitigate Customer Acquisition Cost (CAC) inefficiency, the ten clusters were synthesized into three strategic macro-segments. Promotion tactics are directed towards: (1) Regional Network (25.47%) driven by Word-of-Mouth referrals; (2) Urban Premium (25.00%) via Omnichannel Ads; and (3) Cost-Sensitive Vocational (49.53%) requiring proactive school roadshows and scholarships. This integrated modeling successfully transforms intuitive mass marketing into measurable, efficient, and data-driven precision marketing.]]>
Copyrights © 2026
