"OPTIMIZING CRUDE OIL PRICE PREDICTIONS WITH DEEP LEARNING: A COMPARATIVE PERSPECTIVE"

Abstract

This research investigates the performance of various recurrent neural network (RNN) models—namely Long Short-Term Memory (LSTM), Bidirectional LSTM (Bi-LSTM), Gated Recurrent Units (GRU), and Bidirectional GRU (Bi-GRU)—in the context of time series forecasting for crude oil price prediction. Accurate price forecasting is crucial in financial markets, offering potential benefits for investors and traders by providing valuable insights for decision-making. The study utilizes historical crude oil price data, incorporating open, close, high, and low prices, along with trading volume and other relevant indicators. After preprocessing and splitting the data into training and testing sets, each RNN model is trained to capture the temporal patterns and dependencies inherent in the price movements. The performance of each model is evaluated using key metrics, including mean squared error (MSE), mean absolute error (MAE), and root mean squared error (RMSE), to assess their predictive accuracy on the testing set. The results are compared to determine which RNN architecture provides the most reliable and accurate predictions, contributing valuable insights into the application of deep learning for financial forecasting tasks.