Table 6.
Hunger and satiety classification results on each sensor channel using RF classifier.
| Sensor | Acc. Hungry | Acc. Satiety | Acc | AF1 Score |
|---|---|---|---|---|
| Tmp | 73.08 | 95.30 | 92.00 | 84.19 |
| Eda_E4 | 70.59 | 94.98 | 91.43 | 82.79 |
| BVP | 67.35 | 94.68 | 90.86 | 81.02 |
| Eda_RB | 62.18 | 92.25 | 87.14 | 77.22 |
| IBI | 43.48 | 91.95 | 85.14 | 67.46 |
| HR | 40.45 | 91.33 | 84.86 | 65.89 |
| EMG | 30.95 | 90.58 | 83.43 | 60.77 |
| Resp | 29.30 | 79.56 | 68.29 | 54.43 |
| EOG | 39.25 | 73.25 | 62.86 | 56.25 |
| ECG | 21.59 | 73.66 | 60.57 | 47.63 |
RF: Random forest classifier; Acc: Accuracy; BVP: Blood volume pulse; Eda_E4: Electrodermal activity sensor of empatica E4 wristband; Tmp: Thermopile; IBI: Inter-beat interval; HR: Heart rate; Resp: Respiratory; Eda_RB: Electrodermal activity sensor of RespiBan; ECG: Electrocardiogram; EMG: Electromyography; EOG: Electrooculography. Note: For these experiments, we used a window size of 60 s and a step size of 30 s to compute the 18 hand-crafted features for each axis of the sensor channel.