Table 7 Regularization-Based ablation study for deep learning Architectures.
From: An IoT-enabled AI framework for sustainable product design optimizing eco-efficiency using BiLSTM
Model Variant | Accuracy (%) | Precision (%) | F1-Score (%) | AUC | MCC | Energy (kWh) |
|---|---|---|---|---|---|---|
CNN (Full) | 92.3 | 91.5 | 91.4 | 0.940 | 0.85 | 16.5 |
CNN (No Dropout) | 89.7 | 88.4 | 88.1 | 0.912 | 0.81 | 17.9 |
LSTM (Full) | 91.5 | 90.3 | 91.2 | 0.941 | 0.88 | 14.3 |
LSTM (No Dropout) | 88.9 | 87.1 | 87.6 | 0.914 | 0.83 | 15.6 |
RNN (Full) | 88.4 | 87.0 | 86.8 | 0.905 | 0.82 | 15.1 |
RNN (No Dropout) | 85.6 | 83.9 | 84.2 | 0.876 | 0.77 | 16.4 |
GRU (Full) | 90.1 | 88.7 | 89.8 | 0.929 | 0.86 | 13.7 |
GRU (No Dropout) | 87.6 | 85.9 | 86.1 | 0.901 | 0.81 | 15.0 |
BiLSTM (Full) | 97.6 | 96.2 | 96.1 | 0.980 | 0.93 | 12.5 |
BiLSTM (No Dropout) | 95.1 | 93.8 | 93.4 | 0.955 | 0.89 | 13.9 |
