Financial Contagion Analysis Based on Hybrid Nonlinear Mutual Prediction Algorithm and Fuzzy Neural Networks

Abstract

The contagion of the financial crisis became more and more evident since 1990s. The traditional works are mostly based on the linear methods, which have limitation to investigate the nonlinear features of the financial contagion. The hybrid nonlinear mutual prediction algorithm and fuzzy neural networks method is used in this paper to work on it. The searching of the dynamical interdependence is the first and important step of the financial contagion study. The basis is the multi-series nonlinear mutual prediction method, which means that mutual predictability of two time series is the evidence of the dynamical interdependence. The hybrid nonlinear mutual prediction and fuzzy neural networks are used to detect the dynamical interdependence of the stock indices time series of China, Hong Kong and USA. The results show that there exists strong mutual predictability in each pair of these economies, which means that dynamical interdependence is detectable in these economies.