Research Article
CNN-LSTM-Based Late Sensor Fusion for Human Activity Recognition in Big Data Networks
Table 3
The weighted f1-score for the PAMAP2 dataset using all five models.
| Activity label | Activities | CNN-EF | CNN-LF | LSTM-EF | CNN-LSTM-EF | CNN-LSTM-LF |
| 0 | Lying | 70.22 | 84.97 | 88.66 | 98.63 | 98.25 | 1 | Sitting | 75.88 | 76.04 | 81.62 | 97.81 | 97.32 | 2 | Standing | 85.15 | 93.70 | 96.86 | 95.63 | 93.12 | 3 | Walking | 91.83 | 95.25 | 85.77 | 75.12 | 94.35 | 4 | Running | 92.58 | 94.83 | 92.85 | 95.89 | 97.40 | 5 | Cycling | 92.62 | 95.36 | 88.11 | 94.23 | 96.83 | 6 | Nordic walking | 84.49 | 86.75 | 85.15 | 72.32 | 94.63 | 7 | Ascending stairs | 68.05 | 56.55 | 78.58 | 82.58 | 85.36 | 8 | Descending stairs | 77.66 | 80.70 | 78.91 | 83.76 | 84.81 | 9 | Vacuum cleaning | 86.52 | 90.65 | 85.60 | 88.86 | 86.14 | 10 | Ironing | 88.95 | 89.92 | 94.37 | 94.78 | 91.47 | 11 | Rope jumping | 93.80 | 97.44 | 99.23 | 91.95 | 89.38 |
|
|