<![CDATA[The increasing participation of retail investors in Indonesia has not been accompanied by adequate investment decision-making quality. This issue is reflected in fear of missing out behavior and mismatches between investment choices and risk profiles, particularly among beginner investors. This study develops a personalized stock portfolio recommendation system by integrating K-Means Clustering and Mean-Variance Optimization based on investor risk profiles. The analysis covers 39 liquid LQ45 stocks using four years of historical data. The clustering process identified an optimal K value of 4 with a silhouette score of 0.274, resulting in four stock groups: Low Performers, Value & High Dividend, High ROE, and Aggressive Growth. MVO generated three portfolios with Sharpe ratios between 1.19 and 1.21. The aggressive portfolio achieved an expected return of 24.79% with a volatility of 20.53%. The system was implemented as a web application integrating a BCA-standard risk profiling questionnaire and portfolio weight conversion into exchange-compliant lot units. Unlike a previous Fuzzy C-Means-based approach that used one representative stock from each cluster as MVO input, this study optimizes all 39 stocks and incorporates investor risk profile calibration, producing directly executable portfolio recommendations.]]>
Copyrights © 2026
