Application of K-Medoids Clustering for The Formation of Mean-Semivariance Portfolios on The Idxshagrow Index

Abstract: The number of investors in the Indonesian capital market continues to grow, underscoring the need for investment strategies that deliver optimal returns with controlled risk. Since investors prioritize downside risk over profit fluctuations, this study applies a downside-risk approach to portfolio formation. Stocks in the IDX Sharia Growth (IDXSHAGROW) index are grouped using K-Medoids Clustering based on Return on Assets (ROA) and Return on Equity (ROE). From each cluster, one stock with the highest expected return is selected to construct a Mean–Semivariance portfolio. Portfolio performance is evaluated using the Sharpe Ratio, while risk is measured using the Historical Simulation Value-at-Risk method. The results show that four optimal clusters were formed. The Mean–Semivariance portfolio consists of EMTK (25.16%), RAJA (2.47%), SSIA (9.72%), and TAPG (62.66%), with an expected return of 0.003446. A positive Sharpe Index value of 0.144198 indicates that the portfolio outperforms the risk-free return. VaR risk measurement at a 95% confidence level with a 1-day holding period of IDR 301,801.4, assuming investment capital of IDR 10,000,000.