International Early Warning Nonlinear Model for Securities Systemic Risk Based on RBFNN Using UDM and BVM
Diao Xiao-hua, Kang Shiying · Procedia Engineering · 2012
In the process of the international early warning for securities systemic risk, the uniform design method (UDM) and boundary value method(BVM) are used to design the uniform, large-scale and representative samples to train the RBF neural network (RBFNN). The experiments show, the difference between the real output value of RBFNN early warning model and the expected early warning fussy evaluation value of Experts is very small. The designed early warning method overcomes the difficulties of lack of large-scale and high quality training samples,the lack causing the traditional RBFNN's poor generalization ability. Meantime the methods realizes the non-linear approaching ability, avoids the non-repeatability and instability of traditional experts early warning.