The Chaotic Prediction for Aero-Engine Performance Parameters Based on Nonlinear PLS Regression

Table 1

Comparative analysis of the optimal results based on various forecasting methods.

Forecasting method	Embedding dimension (m)	Delay time (τ)	Subset numbers	Largest Lyapunov exponent	NMSE
Cubic spline interpolation	9	3	5	0.25	0.6400
Uniform kernel	8	19	2	0.16	0.5529
Epanechnikov kernel	8	19	2	0.17	0.5529
Bi-weight kernel	8	19	2	0.38	0.5431
Tri-weight kernel	9	3	5	0.29	0.5918
Gaussian kernel	3	10	4	0.32	0.6325
PLS	9	3	0	0.15	0.6469
OLS	30	16	0	0.098	0.7081