Table 6 Performance comparison of LDA-based feature selection with different deep learning classifiers on the pest dataset * Here feature selection methods are DenseNet201–DN201, EfficientNetB3–ENB3, InceptionResNetV2–INV2 and Multi-features–MF.
Classifiers | BiLSTM | DenseNet | LSTM | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
Feature selection technique | DN201 | ENB3 | INV2 | MF | DN201 | ENB3 | INV2 | MF | DN201 | ENB3 | INV2 | MF |
Accuracy | 98.34 | 99.50 | 99.04 | 99.9 | 89.86 | 97.44 | 92.90 | 99.99 | 93.71 | 98.08 | 94.55 | 99.97 |
Validation accuracy | 86.80 | 95.53 | 89.95 | 100 | 86.90 | 95.84 | 91.17 | 100 | 87.01 | 94.01 | 89.14 | 100 |
Precision | 98.61 | 99.54 | 99.13 | 99.9 | 97.38 | 98.84 | 97.69 | 99.99 | 95.39 | 98.55 | 96.18 | 99.97 |
Validation precision | 91.95 | 96.60 | 92.31 | 100 | 94.45 | 98.38 | 95.65 | 100 | 91.82 | 95.72 | 93.71 | 100 |
Recall | 97.96 | 99.36 | 98.89 | 99.9 | 87.22 | 96.88 | 91.15 | 99.99 | 92.26 | 97.44 | 93.16 | 99.97 |
Validation recall | 84.57 | 95.43 | 89.04 | 100 | 84.06 | 94.62 | 89.65 | 100 | 84.47 | 92.89 | 87.51 | 100 |
Loss | 0.05 | 0.02 | 0.03 | 0.00 | 0.31 | 0.09 | 0.22 | 0.001 | 0.18 | 0.07 | 0.16 | 0.001 |
Validation loss | 0.40 | 0.18 | 0.30 | 0.00 | 0.39 | 0.16 | 0.27 | 0.001 | 0.42 | 0.29 | 0.37 | 0.001 |
