A Combined Model Based on GRU with Mahalanobis Distance for Oil Price Prediction

Oil plays a key role in economic development. It could help people to make plans and decisions. Because oil price is affected by some factors, it is difficult for people to predict it accurately with current existing models. With the development of deep learning, it is often used to solve multivariate nonlinear problems. In this paper, a combined model based on gated recurrent units is proposed, Mahalanobis distance is used to eliminate outliers and the multivariate nonlinear regression model is constructed, Spearman correlation coefficient is used as feather selection and evaluation metric. Experiment shows that the proposed combined method performs better and could predict the price of crude oil effectively.