Using neural networks to tune the fluctuation of daily financial condition indicator for financial crisis forecasting

Recently, Oh et al. [11, 12] developed a daily financial condition indicator (DFCI) which issues an early warning signal based on the daily monitoring of financial market volatility. The major strength of DFCI is that it is expected to serve as a quite useful early warning system (EWS) for the new type of crisis which starts as an instability of the financial markets and then develops into a major crisis (e.g., 1997 Asian crises). One of the problems with DFCI is that it may show a high degree of fluctuation because it handles daily variable, and this may harm its reliability as an EWS. The main purpose of this article is to propose and discuss a way of smoothing DFCI, i.e., it will be tuned using long-term (monthly or quarterly) fundamental economic variables. It turns out that such a tuning procedure could reveal influential macroeconomic variables on financial markets. Since tuning DFCI is done by the method of fitting various types of data simultaneously, neural networks are employed. Tuning the DFCI for the Korean financial market is given as an empirical example.