Table 1.
Clarke Error Grid Analysis for 30- and 60-minute Prediction Horizons for AR Models Based on Subjects #6 and #8a Prediction horizon
| Prediction horizon | ||||||||
|---|---|---|---|---|---|---|---|---|
| 30 min | 60 min | |||||||
| Model-Subject | Model-Subject | |||||||
| Zone | 6-6 | 6-8 | 8-6 | 8-8 | 6-6 | 6-8 | 8-6 | 8-8 |
| A | 85.3% | 84.4% | 82.2% | 90.0% | 66.2% | 64.2% | 60.7% | 72.9% |
| B | 13.3% | 14.2% | 15.0% | 9.8% | 31.1% | 32.5% | 32.9% | 25.1% |
| C | 0.0% | 0.0% | 0.0% | 0.0% | 0.6% | 0.2% | 0.8% | 0.0% |
| D | 1.4% | 1.4% | 2.8% | 0.2% | 2.1% | 3.1% | 5.4% | 2.0% |
| E | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% |
a
For predictions based on models derived from the subject's own data (6-6 and 8-8), the analysis is performed on the last 4000 data points. For portable models, where a model based on subject's #6 data is used to predict subject #8 (6-8), as illustrated in Figure 6, and vice versa (8-6), as illustrated in Figure 7, the analysis is performed over the entire range of the predicted 6000 data points.