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

Table 3

Optimal results of three engines based on various forecasting methods.

Forecasting method	Second aeroengine				Third aeroengine				Fourth aeroengine
	𝑚	𝜏	Subset numbers	NMSE	𝑚	𝜏	Subset numbers	NMSE	𝑚	𝜏	Subset numbers	NMSE
Cubic spline interpolation	8	5	4	0.6041	9	2	5	0.3354	17	3	5	0.3800
Uniform kernel	12	2	4	0.5411	7	2	2	0.3741	10	3	4	0.3529
Epanechnikov kernel	12	2	4	0.5706	9	2	2	0.3417	17	3	4	0.3329
Bi-weight kernel	12	2	4	0.5143	9	2	2	0.2574	17	3	4	0.2431
Triweight kernel	11	3	3	0.5717	9	3	4	0.2805	19	5	5	0.3918
Gaussian kernel	5	9	3	0.5675	14	2	4	0.4521	17	2	3	0.4325
PLS	26	3	0	0.6876	17	3	0	0.4688	28	3	0	0.4192
OLS	12	7	0	0.7515	8	11	0	0.5742	16	7	0	0.4937