Comparison of LSTM, GRU, and CNN Models for Gold Price Forecasting

Objective

Compare LSTM, GRU, and CNN architectures to determine the most effective deep learning approach for gold price forecasting using historical market data from 2020–2024.

Project Details

Analyzed and preprocessed approximately 1,800 daily gold price records (2020–2024) to develop deep learning–based forecasting models.
Designed, trained, and evaluated multiple architectures (LSTM, GRU, and CNN) for time-series prediction, comparing model accuracy, efficiency, and forecasting performance across different prediction horizons.
Identified GRU as the best-performing model, achieving over 90% prediction accuracy while maintaining superior computational efficiency compared to alternative architectures.
Presented research findings at a National Mathematics Seminar, demonstrating the practical application of deep learning techniques for financial market forecasting.

Key Results

✓GRU achieved the best overall performance, delivering over 90% prediction accuracy while maintaining higher training efficiency than LSTM and CNN.
✓Model effectiveness varied across forecasting horizons, emphasizing the importance of selecting an appropriate time window for financial time-series forecasting.

Comments (1)

Rian HidayatJul 21, 2026

Perbandingan model GRU vs LSTM untuk harga emas ini menyajikan analisis komparatif yang tajam.

Project Information

Date:Oct 2024

Duration:2 months

Tools & Technologies

PythonTensorFlowKerasPandasNumPyScikit-learnMatplotlibJupyter Notebook