Table 4 Performance comparison of different prediction models on the test dataset, demonstrating the superior accuracy and coefficient of determination achieved by the proposed LSTM-Attention architecture relative to conventional and alternative deep learning approaches.
From: Digital twin-driven deep learning prediction and adaptive control for coal mine ventilation systems
Model | MAE | RMSE | MAPE (%) | R² | Training Time (min) |
|---|---|---|---|---|---|
LSTM-Attention (Proposed) | 8.23 | 12.45 | 2.87 | 0.9612 | 38.5 |
LSTM | 11.67 | 17.82 | 4.21 | 0.9324 | 32.8 |
GRU | 12.34 | 18.96 | 4.58 | 0.9268 | 29.6 |
CNN-LSTM | 10.52 | 15.73 | 3.74 | 0.9441 | 41.2 |
ARIMA | 24.89 | 35.67 | 8.93 | 0.7856 | 5.3 |
SVR | 19.45 | 28.34 | 7.12 | 0.8472 | 18.7 |
