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

Table 9

Best performance obtained by NFHB-Class¹ (fixed selection strategy), NFHB-Class² (adaptive selection strategy) and neural networks, for rock classes: gneiss and basalt; diabase, and rhyolite; basalt, diabase, and rhyolite; gneiss, basalt, diabase, and rhyolite.

Rock	Model	Correct percentage of training set	Correct percentage of validation set	Number of generated rules
Gneiss and basalt	NFHB-Class^1	96%	96%	53
	NFHB-Class^2	98.67%	96.67%	68
	Neural network	100%	94%	—
Diabase and rhyolite	NFHB-Class^1	81.6%	86%	22
	NFHB-Class^2	87.2%	86%	33
	Neural network	99.2%	85.2%	—
Basalt, diabase, and rhyolite	NFHB-Class^1	84%	80.57%	77
	NFHB-Class^2	84.29%	80.29%	33
	Neural network	97.42%	77.42%	—
Gneiss, basalt, diabase, and rhyolite	NFHB-Class^1	77%	75.75%	105
	NFHB-Class^2	83.75%	77.75%	79
	Neural network	97%	77%	—