Classification Modeling Method for Near-Infrared Spectroscopy of Tobacco Based on Multimodal Convolution Neural Networks

Table 4

Classification performance of 1-D CNN models with different sampling kernel sizes.

Model	Convolution kernel size	Sampling kernel size	Test set classification accuracy (%)
1D_CNN_11	1 ∗ 15	1 ∗ 1	91.14
	1 ∗ 15	1 ∗ 2	92.82
	1 ∗ 15	1 ∗ 3	90.05
	1 ∗ 15	1 ∗ 4	89.53
	1 ∗ 15	1 ∗ 5	88.47