Forecasting crude oil prices with alternative data and a deep learning approach

As crude oil is an essential energy source, fluctuations in crude oil prices are crucial to economic development. Considering the great impact of the COVID-19 outbreak on the financial market, we use the convolutional neural network (CNN) method to forecast oil prices with 24 price-related technical indicators, COVID-19 infections and the Baltic Dry Index (BDI). We further compare its prediction ability with traditional machine learning algorithms, including decision trees, support vector machines, and random forests. We find that the CNN has good forecasting ability both before and after the COVID-19 epidemic. In addition, during the COVID-19 pandemic, the BDI and COVID-19 epidemic-related indicators improved the model forecast accuracy from 2.2 to 10.99%. We show that the CNN could achieve good performance for oil price forecasting during the COVID-19 period..