Crude Oil Cost Forecasting using Variants of Recurrent Neural Network

Abstract

Crude oil cost plays very important role in the country’s economic growth. It is having close impact on economical stability of nation. Because of these reasons it is very important to have accurate oil forecasting system. Due to impact of different factors oil cost data is highly nonlinear and in fluctuated manner. Performing prediction on those data using data driven approaches is very complex task which require lots of preprocessing of data. Working on such a non-stationary data is very difficult. This research proposes recurrent neural network (RNN) based approaches such as simple RNN, deep RNN and RNN with LSTM. To compare performance of RNN variants this research has also implemented Naive forecast and Sequential ANN methods. Performance of all these models are evaluated based on root mean square error(RMSE), mean absolute error(MAE) and mean absolute percentage error(MAPE). The experimental result shows that RNN with LSTM is more accurate compare to all other models. Accuracy of LSTM is more than 96% for the dataset of U.S. Energy Information administration from March 1983 to June 2022. On the basis of experimental result, we come to the conclusion that RNN with LSTM is best suitable for time series data which is highly nonlinear.