TABLE 5.

Ablation experiment.

Network structure	ACC	AUC	PRE	REC	F1	Parameter
CapsNet	0.9304	0.9217	0.9123	0.8956	0.8960	266 304
IR‐caps	0.9778	0.9750	0.9672	0.9667	0.9666	79 515 104
IRC‐caps	0.9851	0.9833	0.9780	0.9778	0.9777	81 876 722
IRCM‐Caps	0.9911	0.9900	0.9866	0.9867	0.9866	57 233 140

Note: CapsNet is the basic network; IR‐caps adds a convolutional neural network on the basis of CapsNet; IRC‐caps adds CBAM module on the basis of IR‐caps; IRCM‐Caps adds MBL on the basis of IRC‐caps Module.; bold font is the optimal value for each column.

Abbreviation: AUC, area under the curve.