Table 6 Performance comparison of ML models (this work) for chipless RFID–IoT environmental sensing.
From: Edge machine learning over IoT for chipless RFID environmental sensing in smart agriculture
Model | Application | Accuracy/MAE | Precision | Recall | F1-score | Training time (s) | Noise resilience (±) | Remarks/Notes |
|---|---|---|---|---|---|---|---|---|
Random forest | Temperature classification (0–90 °C, 10 °C bins) | 96.2% | 0.96 | 0.95 | 0.96 | 3.4 | 10 MHz (jitter) | High accuracy, fast training; ideal for on-edge deployment |
SVR (RBF Kernel) | Continuous temperature regression | ± 1.3 °C (MAE) | – | – | – | 8.1 | 8 MHz (jitter) | Smooth high-precision output; suitable for fine thermal mapping |
XGBoost | Humidity classification (0–90% RH) | 94.5% | 0.95 | 0.94 | 0.95 | 2.8 | 15% RH (amplitude loss) | Excellent discrete-level prediction; lightweight inference |
Gradient boosting regressor | Continuous humidity regression | ± 2.1% RH (MAE) | – | – | – | 5.6 | 12% RH (variation) | Stable under humidity non-linearity; low latency at edge |
k-Means + XGBoost Hybrid | Anomaly detection/spectral drift identification | 98% (AUC) | 0.97 | 0.98 | 0.98 | 4.2 | 20 MHz (jitter equiv.) | Detects multi-tag interference; robust to environmental distortion |
