Optimal cooperation between external criterion and data division in GMDH
Chao He, Jinjie Wu, Johann-Adolf Müller · International Journal of Systems Science · 2008
The problem of cooperation between external criterion and data division is important for application and development of group method of data handling (GMDH) theory. Based on research of ‘critical noise level’ and ‘noise immunity’, this article presents the concept of ‘expected critical noise level’. The relationship between ‘expected critical noise level’ and both prediction set and model structure is expressed by a formula. According to ‘expected critical noise level’, a selection method is designed, which is able to obtain an optimal cooperation between external criterion and data division in GMDH. Finally, the corresponding algorithm and an example are given.