Energy futures price prediction and evaluation model with deep bidirectional gated recurrent unit neural network and RIF-based algorithm

Highlights: • A novel model (RIF-DBGRUNN) is proposed to forecast crude oil futures price. • Deep bidirectional learning and random inheritance formula are combined with model. • Basic metrics and q-DSCID synchronization evaluation is used to measure accuracy. • Results show the RIF-DBGRUNN model outperforms the comparison models. Energy resources have firmly occupied an unshakable position, which is indispensable both in industrial field and daily life. More accurate prediction of energy futures price has always been a challenging issue. Motivated by this problem, a novel random deep bidirectional gated recurrent unit neural network is constructed to achieve more accurate forecasts of international crude oil futures prices. The random inheritance formula is proposed and integrated into the training process of the model, and it reflects the timeliness of historical data. Both the random inheritance formula and the deep bidirectional learning can effectively improve the model's acquisition of effective information from historical data and improve the model's accuracy. The proposed model is compared with SVM, GRU, ERNN, LSTM, DBGRUNN and RIF-GRUNN models, and a variety of evaluation indicators as well as a novel synchronization evaluation method of q-DSCID are used to measure accuracy. The empirical research results of four crude oil futures prices and coarse-grained moving absolute returns show that the proposed model outperforms the comparison models. For the Brent crude oil futures price prediction, its metrics R², MAE, TIC, RMSE and SMAPE are 0.998, 0.200, 0.002, 0.267 and 0.283, which are the best in the comparison models.