Vector quantized radial basis function neural networks with embedded multiple local linear models for financial prediction

Authors

Tony Jan, University of Technology, Sydney
Maria H. Kim, University of Technology, SydneyFollow

RIS ID

38826

Publication Details

Jan, T. & Kim, M. H. (2005). Vector quantized radial basis function neural networks with embedded multiple local linear models for financial prediction. IEEE International Joint Conference on Neural Networks (IEEE-IJCNN) (pp. 2538-2543). Piscataway, NJ: Institute of Electrical and Electronics.

Abstract

In this paper, a model is proposed which combines multiple local linear models with a novel modified probabilistic neural network (MPNN). The proposed model is developed to approximate multiple nonlinear model with reduced computational requirement. The proposed model shows to provide both low bias and variance with reduced computations by utilizing semiparametric local linear approximation and efficient vector quantization of data space. The proposed model is shown to provide comparable performance to other state-of-the-art models in terms of bias, variance and computational requirement in short-term financial prediction.