Advances, Systems and Applications
Table 6 Comparison of performance of five models according to five scenarios: The number of data points is increased from 180k up to 1.6 Million. As for Fog computing facility: Intel Core i3-8100 CPU 3.60GHz at 4GB memory, and the OS is Ubuntu 20.04.1 LTS. BOSS VS model shows excellent scalability while keeping the highest accuracy among the other models. N/A denotes that simulation data is unavailable due to computation crashes caused by hardware limitations in this Fog computing facility. The 10-fold cross-validation accuracy is shown in the 95% confidence level
Scenario | Data points (data sets) | Model | Time (sec) | Accuracy (non CV) | Cross Validation Accuracy |
|---|---|---|---|---|---|
I | 180,000 (300) | WEASEL MUSE | 0.39 | 0.20 | \(0.46 \pm 0.17\) |
BOSS VS | 0.46 | 1.00 | \(0.98 \pm 0.04\) | ||
RF | 0.44 | 0.56 | \(0.70 \pm 0.12\) | ||
LR | 0.25 | 0.63 | \(0.57 \pm 0.14\) | ||
1-NN DTW | 10.40 | 0.26 | \(0.35 \pm 0.06\) | ||
II | 540,000 (900) | WEASEL MUSE | 1.00 | 0.33 | \(0.49 \pm 0.08\) |
BOSS VS | 1.42 | 0.97 | \(0.98 \pm 0.02\) | ||
RF | 1.16 | 0.82 | \(0.85 \pm 0.06\) | ||
LR | 2,700 | 0.65 | \(0.64 \pm 0.06\) | ||
1-NN DTW | 89.54 | 0.37 | N/A | ||
III | 900,000 (1,500) | WEASEL MUSE | 1.810 | 0.34 | \(0.56 \pm 0.08\) |
BOSS VS | 2.456 | 0.98 | \(0.98 \pm 0.01\) | ||
RF | 1.961 | 0.90 | \(0.89 \pm 0.06\) | ||
LR | 7.153 | 0.74 | \(0.70 \pm 0.09\) | ||
1-NN DTW | 246.794 | 0.40 | N/A | ||
IV | 1,080,000 (1,800) | WEASEL MUSE | 2.193 | 0.36 | \(0.54 \pm 0.06\) |
BOSS VS | 2.966 | 0.98 | \(0.98 \pm 0.01\) | ||
RF | 2.472 | 0.87 | \(0.89 \pm 0.03\) | ||
LR | 12.921 | 0.76 | \(0.70 \pm 0.05\) | ||
1-NN DTW | 352.067 | 0.41 | N/A | ||
V | 1,620,000 (2,700) | WEASEL MUSE | 3.851 | 0.37 | \(0.55 \pm 0.03\) |
BOSS VS | 8,100 | 0.98 | \(0.98 \pm 0.01\) | ||
RF | 3.910 | 0.92 | \(0.91 \pm 0.02\) | ||
LR | 32.183 | 0.71 | \(0.73 \pm 0.02\) | ||
1-NN DTW | 793.22 | 0.38 | N/A |