Reinforcement Learning Untuk Optimasi Sharpe Ratio Pada Perdagangan Saham Indonesia

Taufiq, A Muhammad Yusuf (2025) Reinforcement Learning Untuk Optimasi Sharpe Ratio Pada Perdagangan Saham Indonesia. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6032231109-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Pasar saham termasuk Indonesia memiliki tingkat volatilitas yang tinggi akibat pengaruh berbagai faktor termasuk makroekonomi dan geopolitik. Dalam merespons dinamika tersebut, investor umumnya mengandalkan analisis teknikal dan fundamental untuk mengambil keputusan investasi. Namun, Seiring berkembangnya teknologi kecerdasan buatan, reinforcement learning menawarkan pendekatan yang baru yang adaptif dan optimal melalui pembelajaran dari data historis dan umpan balik dari hasil keputusan perdagangan sebelumnya.
Penelitian ini bertujuan untuk menyelidiki penerapan reinforcement learning dalam mengoptimalkan strategi perdagangan saham di Indonesia dengan fokus utama peningkatan Sharpe Ratio, yaitu rasio pengembalian terhadap risiko. Dengan demikian, penelitian ini tidak hanya mengejar keuntungan maksimal, tetapi juga mempertimbangkan pengelolaan risiko yang efektif, sehingga dapat menciptakan strategi yang seimbang antara potensi profit dan stabilitas kinerja investasi.
Metodologi yang digunakan dalam penelitian ini mencakup desain environment RL dengan fitur teknikal pasar saham yang dikonstruksi dari data timeseries mentah Open High Low Close Volume (OHLC) dari 7 saham yang tergabung indeks LQ45 selama 10 tahun terakhir (01 Januari 2015-31 Desember 2024). Algoritma RL yang digunakan antara lain Proximal Policy Optimization (PPO), Deep Deterministic Policy Gradient (DDPG), dan Advantage Actor-Critic (A2C), dengan action space yang terdiri dari aksi beli, jual, dan tahan. Model juga mempertimbangkan beberapa kendala seperti batasan modal, risiko, frekuensi transaksi, dan alokasi dana untuk memastikan strategi yang dihasilkan tetap realistis dan layak diterapkan.
Hasil penelitian ini menunjukkan bahwa reinforcement learning dengan algoritma PPO, A2C, dan DDPG masing-masing mencapai Sharpe Ratio 0.98, 0.76, dan 0.90, dengan annual return 20.95%, 22.47%, dan 28.00%, serta MDD 8.24%, 17.64%, dan 16.20%. Temuan ini mengindikasikan bahwa pendekatan RL mampu menjaga keseimbangan antara return, risiko, dan stabilitas portofolio.
=========================================================================================================================================
The Indonesian stock market, like many others, exhibits high volatility driven by macroeconomic and geopolitical factors. Traditionally, investors rely on technical and fundamental analysis to guide their decisions; however, with advances in artificial intelligence, reinforcement learning (RL) offers a novel and adaptive approach by learning from historical data and feedback from past trading outcomes.
This study aims to investigate the application of RL in optimizing stock trading strategies in Indonesia, with a primary focus on improving the Sharpe Ratio, which measures return relative to risk. The objective is not only to maximize profits but also to ensure effective risk management, thereby creating strategies that balance potential gains with portfolio stability.
The methodology involves designing an RL environment using stock market technical features constructed from raw OHLCV (Open, High, Low, Close, Volume) time series data of seven LQ45 stocks over the past ten years (January 1, 2015 – December 31, 2024). The RL algorithms applied include Proximal Policy Optimization (PPO), Deep Deterministic Policy Gradient (DDPG), and Advantage Actor-Critic (A2C), with an action space consisting of buy, sell, and hold decisions. The model also incorporates constraints such as capital limits, risk thresholds, trading frequency, and fund allocation to ensure that the strategies remain realistic and applicable in practice.
The results show that RL, through PPO, A2C, and DDPG, achieved Sharpe Ratios of 0.98, 0.76, and 0.90, with annual returns of 20.95%, 22.47%, and 28.00%, and maximum drawdowns (MDD) of 8.24%, 17.64%, and 16.20%, respectively. These findings indicate that RL can effectively balance return, risk, and portfolio stability, demonstrating its potential as a robust alternative to classical trading strategies in the Indonesian stock market.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	reinforcement learning, perdagangan saham, stock trading, sharpe ratio
Subjects:	H Social Sciences > HG Finance > HG4915 Stocks--Prices
Divisions:	Interdisciplinary School of Management and Technology (SIMT) > 61101-Master of Technology Management (MMT)
Depositing User:	Muhammad Yusuf Taufiq
Date Deposited:	29 Jan 2026 04:07
Last Modified:	29 Jan 2026 04:07
URI:	http://repository.its.ac.id/id/eprint/130693

Actions (login required)

View Item