<![CDATA[This study aims to develop a classification model using the random forest approach to identify mean reversion events in stock prices within the Indonesian capital market, which applies a long-only trading system and prohibits short selling. A mean reversion event is defined as a condition in which the price deviates from the average—specifically, the 50-period Simple Moving Average (SMA)—and subsequently moves back toward the average within a certain timeframe. The model is trained on historical data from three of the most liquid stocks in the financial sector: BBCA, BMRI, and BBNI. The features used include technical indicators such as SMA, Relative Strength Index (RSI), oscillators, and price autocorrelation. Two models were developed with variations in parameters including the reversal window, number of estimators, maximum depth, class weight, threshold, and classification probability. Evaluation was conducted using the Receiver Operating Characteristic (ROC) curve and precision-recall metrics, and further tested on out-of-sample data from cross-sector stocks to assess the model’s generalization capability across various market conditions. In addition, the buy-sell strategy was tested through simulations and validated using Monte Carlo methods to evaluate the model’s robustness in actual trading conditions and under random data variations. The results indicate that mean reversion events can be effectively modeled, yielding high reliability—particularly in models that are more selective or less responsive to short-term price fluctuations. The model also demonstrated strong simulation performance, especially when implemented with appropriate filtering methods.]]>
Copyrights © 2025
