A Hybrid Machine Learning Approach for Volatility Forecasting Combining GARCH and LSTM

In this paper, a hybrid model is presented that is built using the sGARCH model and the LSTM neural network to predict the volatility of Brent crude oil prices. The main objective of this paper is to compare the accuracy of traditional sGARCH models, machine learning-based LSTM models, and hybrid models in estimating market volatility and to select the best model for forecasting future volatility. Predictions are made for the next ۷۷۴ days. The models are trained on crude oil data from January ۲۰۱۰ to ۲۰۲۲, and their performance is tested using data from ۲۰۲۲ to January ۲۰۲۵ with MAE and MSE error metrics. The results show that the sGARCH model does not perform as well as machine learning models. The hybrid LSTM-sGARCH model, which has the lowest error, is chosen as the best. This suggests that combining the LSTM model’s ability to learn nonlinear patterns with how the sGARCH model handles conditional volatility can lead to more accurate predictions.

کلیدواژه ها:

Machine Learning ، Hybrid Model ، LSTM Model ، sGARCH Models ، Forecast Volatility