Table 6. Unsupervised CEERM recognition performance based on different feature layers.

Model+Abstract Layers	R	P	F	FI	FIR(%)
DBN4+POSL	78.13	66.32	71.74	0	0
DBN4+POSL+DPL	92.42	66.52	77.36	5.62	7.83
DBN4+POSL+DPL+LENL	83.87	77.42	80.52	3.16	4.08
DBN4+POSL+DPL+LENL+LOCL	83.87	78.39	81.04	0.52	0.65
DBN4+POSL+DPL+LENL+LOCL+DISL	80.01	86.32	83.05	2.01	2.48
DBN4+POSL+DPL+LENL+LOCL+DISL+TWFL	87.32	83.12	85.17	2.12	2.56

Note: F_c is the F measure when the number of feature layers is c, FI is difference between F_c and F_c−1,

FIR is equal to $\frac{F_c-F_{c-1}}{F_{c-1}}$.