A Comparative study of Average-Based FTSMC, GBM, and LSTM for JCI Forecasting

Syifa Aulia (Applied Mathematics, IPB University, Bogor, Indonesia)
I Wayan Mangku (Mathematics, IPB University, Bogor, Indonesia)
I Gusti Putu Purnaba (Mathematics, IPB University, Bogor, Indonesia)

Zero : Jurnal Sains, Matematika, dan Terapan
Vol 10, No 2 (2026): Zero: Jurnal Sains Matematika dan Terapan

Stocks are high-risk investment instruments, so a model is needed to assist investors in decision-making. The Jakarta Composite Index (JCI) is a key indicator that measures the performance of all stocks on the Indonesian stock exchange. This study compares the performance of the Average-Based FTSMC, GBM, and LSTM models in predicting JCI movements. The research data consists of daily stock index values from June 1, 2024, to June 29, 2026. The models were compared using the evaluation metrics MAPE, RMSE, and MAE. The results show that the Average-Based FTSMC model provides the highest accuracy (MAPE: 0.925%, RMSE: 86.753, and MAE: 65.162). These findings indicate that the fuzzy time series approach is more adaptive to JCI fluctuations than both classical stochastic models and deep learning models. This study is limited to historical data using a single input variable; therefore, generalizing the results to market conditions requires further investigation.