A Feature Fusion Method with Guided Training for Classification Tasks

Table 3

The accuracy and other performance evaluation indexes of models with fused data and separate data.

Model	Class	TP	FP	FN	Precision	Recall	F1-score	Accuracy
FGT-Net (fused data)	Japanese Spitz	40	6	6	0.870	0.870	0.870	0.878
	Pomeranian	27	6	0	0.818	1.000	0.900
	Samoyed	71	9	13	0.888	0.845	0.866
	Husky	78	9	11	0.897	0.876	0.886
CNN (ShuffleNetv2) (only image data)	Japanese Spitz	18	9	28	0.667	0.391	0.493	0.728
	Pomeranian	15	8	12	0.652	0.556	0.600
	Samoyed	63	38	21	0.623	0.750	0.681
	Husky	83	12	6	0.873	0.933	0.902
DNN (only structured data)	Japanese Spitz	45	0	1	1.000	0.978	0.989	0.780
	Pomeranian	27	1	0	0.964	1.000	0.982
	Samoyed	46	15	38	0.754	0.548	0.634
	Husky	74	38	15	0.661	0.831	0.736