Title: Predicting Indian basket crude prices through machine learning models - a comparative approach
Authors: Pradip Kumar Mitra; Charu Banga
Addresses: Vivekanand Education Society's Institute of Management Studies and Research, 495-497 Collectors Colony, Chembur, Mumbai, 400074, India ' Curtin University, Dubai, Block 11, Fourth Floor, Dubai International Academic City, Dubai, UAE
Abstract: Forecasting crude price can bring some stability in the decision making process for the firms dealing with it. Crude oil is a very volatile commodity so only linear time series modelling is not sufficient to predict its price. A nonlinear model like an artificial neural network is a better choice. The paper tries to test the prediction accuracy of a conventional neural network model and deep learning model using monthly data of Indian basket price of crude oil for 18 years. A simple MLP neural network model and a deep learning model of long short-term memory are used in the present study to find accuracies in predicting the crude price. The paper finds that a simple MLP model can provide better forecasting accuracy compared to a complicated LSTM model.
Keywords: crude oil price; forecasting; machine learning models; multilayer perceptron; MLP; neural network; long short-term memory; LSTM.
International Journal of Business Forecasting and Marketing Intelligence, 2019 Vol.5 No.3, pp.249 - 266
Received: 08 Mar 2019
Accepted: 17 Apr 2019
Published online: 09 Dec 2019 *