Prediksi Harga Emas Harian Dan Multi-Horizon Menggunakan ARIMA, LSTM Dan Model Hybrid ARIMA-LSTM

Abstract

Gold price forecasting is a crucial task in financial analysis due to high market volatility and dynamic price movements. This study aims to predict daily gold prices and perform multi-horizon forecasting up to 10 days ahead using Autoregressive Integrated Moving Average (ARIMA), Long Short-Term Memory (LSTM), and a hybrid ARIMA-LSTM model. The dataset consists of daily gold closing price data from January 2022 to December 2024 obtained from Yahoo Finance. Data preprocessing includes handling missing values, normalization using Min–Max scaling, and stationarity testing. ARIMA and LSTM models are developed independently to capture linear and nonlinear patterns, respectively, while the hybrid model combines both approaches by modeling ARIMA residuals using LSTM. Experimental results show that the hybrid ARIMA-LSTM model achieves the lowest prediction error compared to individual models, as indicated by RMSE, MAE, and MAPE values. Furthermore, multi-horizon forecasting results demonstrate that the hybrid model provides more stable and accurate short-term gold price predictions. These findings confirm that the hybrid modeling approach is effective and can support investors and analysts in decision-making processes.