Macroscopic Rock Texture Image Classification Using a Hierarchical Neuro-Fuzzy Class Method

Table 5

Best performance obtained by methods: NFHB-Class¹ (fixed selection strategy), NFHB-Class² (adaptive selection strategy) and neural networks. Rock subclasses: gneiss, basalt, diabase, and rhyolite.

Rock	Model	Best performance for the training set	Best performance for the validation set	Number of generated rules
Gneiss	NFHB-Class^1	100%	98%	52
	NFHB-Class^2	100%	98%	12
	Neural network	100%	96%	—
Basalt	NFHB-Class^1	95%	87%	81
	NFHB-Class^2	88%	84%	25
	Neural network	100%	86%	—
Diabase	NFHB-Class^1	81.6%	71.2%	110
	NFHB-Class^2	92%	73.6%	83
	Neural network	93.6%	69.6%	—
Rhyolite	NFHB-Class^1	93.6%	78.4%	225
	NFHB-Class^2	96%	78.4%	56
	Neural network	97.6%	75.2%	—