<![CDATA[Stock portfolio management in emerging markets such as Indonesia remains challenging due to high volatility, market inefficiencies, and the strong presence of retail investors. In this setting, conventional approaches, including buy-and-hold strategies, the Markowitz framework, and the Capital Asset Pricing Model (CAPM), often struggle to perform consistently under rapidly changing market conditions. While reinforcement learning (RL) has gained increasing traction in global finance, its application in the Indonesian stock market remains limited. This study examines the effectiveness of an RL-based approach, specifically the Deep Q-Network (DQN) algorithm, in optimizing stock portfolios on the Indonesia Stock Exchange (IDX). Using a quantitative experimental design, the analysis is based on back-testing simulations of IDX30 stocks over the 2022–2024 period, with samples selected purposively based on liquidity and market capitalization. The findings show that the DQN-based strategy consistently outperforms conventional methods, delivering higher returns, improved risk–return efficiency, and better control of downside risk. These results suggest that RL models are better suited to adapt to dynamic market conditions. Theoretically, this study extends portfolio optimization literature by incorporating adaptive, learning-based models into emerging market contexts. Practically, it offers evidence for investors and practitioners to consider AI-driven strategies as a more responsive alternative to traditional approaches in a volatile market.]]>
Copyrights © 2026
