IDX30 Portfolio Construction using K-Means Clustering with MAD Risk Optimization and Sortino Ratio Evaluation

Abstract

A stock portfolio plays an important role in managing risk and achieving optimal returns in volatile markets. This study proposes an integrated framework that combines K-Means Clustering, Mean Absolute Deviation (MAD), and the Sortino Ratio. The main contribution lies in linking clustering-based asset selection with downside risk optimization and evaluation, enabling portfolio construction that accounts for asset similarity, risk measurement, and investor-oriented performance assessment. This approach addresses the limitation of previous studies that apply these methods separately by providing a more structured basis for downside risk-adjusted portfolio selection. Using daily IDX30 stock data from June 2024 to June 2025, samples were selected based on index consistency. The results indicate that a portfolio of ANTM and INDF achieved the highest Sortino Ratio of 0.2179. These findings suggest that combining high-return stocks with moderate volatility, supported by clustering and downside risk optimization, can improve downside risk-adjusted performance, providing practical guidance for investors.