<![CDATA[This paper presents a hybrid ARIMA–LSTM model to forecast daily gold price using historical data from Yahoo Finance. Gold price is highly volatile due to macroeconomic, geopolitical, and monetary factors, making accurate forecasting difficult and increasing uncertainty in investment decisions. In this study, ARIMA is used for modeling linear patterns in the time series data, while an LSTM network captures the nonlinear relationships and temporal dynamics that are not captured by statistical models. The dataset consists of daily observations of gold prices between June 2022 and June 2025. The analysis involves cleaning and normalizing the data, splitting it into training and testing subsets, estimating ARIMA parameters, extracting residuals, and forecasting these residuals with LSTM. Performance evaluation is carried out through MAE, RMSE, and MAPE metrics. The hybrid framework compares favorably against standalone ARIMA and LSTM models in terms of all three metrics used for assessment. Empirical results show that the hybrid ARIMA–LSTM model produces lower forecasting errors than the individual models on all evaluation metrics. These findings validate that combining statistical time series modeling with neural sequence learning increases predictive reliability in volatile commodity markets. The proposed framework can be considered methodologically sound for gold price forecasting and subsequently may enhance informed decision-making within financial analysis as well as investment practice.]]>
Copyrights © 2025
