Predictive Analysis of Stock Prices Using ARIMA and Hybrid ARIMA-LSTM Models in the Indonesian Market

This paper presents a hybrid ARIMA-LSTM approach to predict stock prices in the Indonesian market and compares its performance to the standalone ARIMA model. The study focuses on six companies representing major sectors, including finance, energy, industry, primary, health, and non-primary sectors. Using historical data from 2014 to 2023, the research spans three economic phases: pre-pandemic, pandemic, and post-pandemic. The hybrid model leverages residual-based modeling to address high volatility and white noise. Extensive hyperparameter tuning and repeated testing ensure the model’s predictions are stable and reliable. The Mean Absolute Percentage Error ranges from 0.94% to 1.96%, demonstrating the hybrid model’s superior adaptability and accuracy across diverse market conditions.