A Novel Feature Extraction-based Selective & Nonlinear Neural Network Ensemble Model for Economic Forecasting

In this study, a novel selective & nonlinear neural network ensemble model, i.e. NSNNEIPCABag, is proposed for economic forecasting. In this model, some different training subsets are first generated by bagging algorithm. Then the feature extraction technique, improved principal component analysis (IPCA), and then the IPCA approach is also used to extract their data features to train individual networks, and to select the appropriate number of ensemble members from the available networks. Finally, the selected members are aggregated into a nonlinear ensemble model with support vector regression (SVR). For illustration and testing purposes, the proposed ensemble model is applied for economic forecasting.