Prediksi Harga Emas Menggunakan Algoritma Long-Short Term Memory dengan Optimasi Adaptive Momen Estimation

Abstract

Accurate gold price prediction is a crucial aspect of investment decision-making. This study implements a Long Short-Term Memory (LSTM) model, optimized with Adaptive Moment Estimation (Adam), to forecast daily gold prices based on historical data from the 2000-2024 period. The experimental results demonstrate the superior performance of the LSTM-Adam model, which achieved high accuracy with a Mean Absolute Percentage Error (MAPE) of 0.96% and a Mean Squared Error (MSE) of 320,854,198. This model proved to outperform benchmark models such as GRU, ARIMA, and Random Forest by exhibiting the lowest error rates. Furthermore, projections for the next year indicate a consistent upward trend in prices. These findings confirm that the LSTM-Adam architecture is a reliable and robust approach for gold price forecasting, offering a practical contribution to investment strategies and opening opportunities for further applications in modeling other commodity prices.