表 4.

不同特征融合方式的实验结果对比

Experimental results of different feature fusion methods

Models	Methods	Accuracy (%)	Precision (%)	Recall (%)	F1-score (%)
cat: Using concatenation to fuse features. w_cat: Using weighted concatenation to fuse features. w_add: Using weighted addition to fuse features. w_mul: Using element-wise multiplication to fuse features.
OM&IM&TEM model	cat	87.92±4.25	88.67±3.88	88.13±4.44	87.93±4.20
	w_cat	88.34±2.12	88.73±2.40	88.27±2.23	88.13±2.08
	w_add	88.34±2.83	89.13±2.22	88.40±3.07	88.13±2.75
	w_mul	85.83±6.10	86.53±6.14	86.40±5.96	85.87±6.03