Pengembangan Sistem Prediksi Saham Menggunakan Model Hybrid Gated Recurrent Unit–Long Short-Term Memory Berbasis Integrasi Indikator Teknikal Konvensional

Aryanto, Fajar Hanggoro Dwi
Sejati, Rr. Hajar Puji
Sanjaya, Fadil Indra

TIN: TERAPAN INFORMATIKA NUSANTARA
Vol 6 No 8 (2026): January 2026

Stock price prediction is a crucial aspect of investment decision-making in the Indonesian capital market. This study aims to design a hybrid Gated Recurrent Unit–Long Short-Term Memory (GRU–LSTM) model architecture integrated with technical indicators such as Moving Average Convergence Divergence, Moving Average, Exponential Moving Average, and Relative Strength Index to improve the accuracy and objectivity of predictions. Additionally, this study aims to optimize model performance through grid search and implement it into a Flask-based web application as a decision support system for investors. The system was developed using a research and development approach at the Yogyakarta University of Technology. Historical data on PT Bank Rakyat Indonesia (Persero) Tbk. (BBRI.JK) shares for the period from January 2, 2020, to October 17, 2025, was obtained through the Yahoo Finance API as the main dataset. The model was optimized to determine the best combination of hyperparameters. Evaluation was performed using the Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE) metrics. The test results show that the model achieved MAE 0.0241, MSE 0.0012, RMSE 0.0346, and MAPE 2.7%, indicating a high level of accuracy. The web application provides interactive visualization dashboard features, model development, and educational documentation. These findings confirm that the integration of deep learning with technical indicators is an effective solution for more measurable and systematic stock analysis.