Using hybrid Artificial Neural Network with Supportive Vector Regression Technique in Modeling and Prediction/With Practical Application
Keywords:support vector regression (SVR), loss function, kernel function, artificial neural networks, nodes Neurons, perceptron network, multi-layer perceptron networks.
The forecasting process in time series is directly affected by the selection of the appropriate model for the time series data, as this step directly affects the accuracy of the obtained predictions. In this study, a new, unconventional mechanism is used to predict oil prices in Iraq by using the hybridization of the support vector regression with artificial neural networks to predict Iraqi oil prices. A hybrid ANN neural network was built by inserting the svr model as one of the network inputs. The study found the superiority of the ANN (yt,t,svr) network model with two hidden layers, the first is three nodes and the second is one node. According to the training data error , error of net testing, accuracy of net training and accuracy of net testing. The selected network was used to predict the price of Iraqi oil for the next ten years.