<![CDATA[This research examines how machine learning techniques, specifically Random Forest and Long Short-Term Memory (LSTM) models enhanced by technical analysis indicators, can effectively predict stock price, particularly downturn to manage risks in government stock investment. Given the high-risk, high-return profile of stocks and their limited use in government strategies where legal complications is intense, it is essential to promptly identify potential declines in stock prices to mitigate losses and prevent legal issues. This research uses all stock during 2000-2022 from the Indonesian capital market and focuses on integrating predictive models with trading strategies such as active trading guided by predictions, delayed trading decision to account for administrative processes, and drawdown-based strategies to mainly minimizing risks while keeping potential returns optimum. The univariate LSTM model with a 7-day lag (n_lag 7) exhibited the best overall performance, followed by the Random Forest model. Daily active trading strategies were most effective when using LSTM multivariate models, with a 5% drawdown limit proving optimal for managing risks while maximizing returns.]]>
Copyrights © 2024
