A Combined Model Based on Recurrent Neural Networks and Graph Convolutional Networks for Financial Time Series Forecasting

Accurate and real-time forecasting of the price oil plays an important role in world economy. Research interest this type time series has increased considerably recent decades, since, due to characteristics series, it was a complicated task with inaccurate results. Concretely, deep learning models such as Convolutional Neural Networks (CNNs) Recurrent (RNNs) have appeared field promising results compared traditional approaches. To improve performance existing networks forecasting, work two types neural are brought together, combining Graph Network (GCN) Bidirectional Long Short-Term Memory (BiLSTM) network. This is novel evolution that improves literature provides new possibilities analysis series. The confirm better combined BiLSTM-GCN approach BiLSTM GCN separately, well models, lower error all metrics used: Root Mean Squared Error (RMSE), (MSE), Absolute Percentage (MAPE) R-squared (R2). These represent smaller difference between result returned by model real value and, therefore, greater precision predictions model.