Soft Sensor Development Based on Quality-Relevant Slow Feature Analysis and Bayesian Regression with Application to Propylene Polymerization

Table 2

Quality prediction results of different models on the testing dataset.

Model	MAE	MRE (%)	RMSE	TIC	STD
LASSO	0.0187	0.71	0.0233	0.0045	0.0212
RVR	0.0174	0.66	0.0219	0.0042	0.0217
QSFR	0.0168	0.64	0.0189	0.0036	0.0139
PCA-RVR	0.0119	0.45	0.0144	0.0027	0.0126
QSFA-RVR	0.0051	0.19	0.0069	0.0013	0.0069