Table 10 Evaluating CNN with feature selection using 80–20 splitting method using different parameters values
From: Classification of lung cancer severity using gene expression data based on deep learning
Dataset | Convolutional layers | Max-pooling layer | Max-pooling layer size | Dense layer | Learning rate | Batch size | Accuracy |
|---|---|---|---|---|---|---|---|
LUAD using 5000 features | 1 | NA | NA | 1 | 0.001 | 32 | 88.89% |
0.01 | 32 | 89.90% | |||||
16 | 89.90% | ||||||
64 | 89.90% | ||||||
128 | 93.94% | ||||||
256 | 91.92% | ||||||
3 | 3 | 1 | 3 | 0.01 | 128 | 91.92% | |
3 | 0.01 | 75.76% | |||||
1 | 0.001 | 81.82% | |||||
LUSC using 3000 features | 1 | NA | NA | 1 | 0.001 | 32 | 85.26% |
0.01 | 32 | 82.11% | |||||
16 | 85.26% | ||||||
64 | 84.21% | ||||||
128 | 88.42% (training time = 5.7 Sec) | ||||||
256 | 88.42% (training time = 5.6 Sec) | ||||||
3 | 3 | 1 | 3 | 0.01 | 128 | 75% | |
2 | 0.01 | 128 | 78% | ||||
1 | 0.001 | 128 | 75% |