Intelligent Ensemble Systems for Modeling NASDAQ Microstructure: A Comparative Study

摘要

In this study, four neural networks (NN) ensemble systems are presented and compared for NASDAQ returns prediction. They are the conventional feed-forward back-propagation neural network (FFNN) ensemble which widely used in the literature, time-delay neural network (TDNN) ensemble, nonlinear auto-regressive with exogenous inputs (NARX) ensemble and the radial basis neural network (RBFNN) ensemble. Each component of the NN ensemble is used to learn specific patterns related to a given NASDAQ submarket. Based on the mean of absolute errors (MAE), the experiments show that ensemble models based on advanced NN architectures such as TDNN, NARX, and RBFNN ensemble all achieve lower forecasting errors than traditional FFNN ensemble system. In addition, the RBFNN ensemble outperformed all other NN ensembles under study.