From: An integrated priority-based cell attenuation model for dynamic cell sizing
Experiment | Relative error (range) | % Error (%) | % Error at peaks (%) | Rate of calls dropped (starting) | Rate of calls dropped (ending) | |
---|---|---|---|---|---|---|
A1 | Training of previous half-year’s data | 0 to 2.8 × 104 | 0–120 | 7800 | 3900 | 0–120 |
A2 | Prediction of subsequent half-year’s data | 0–2.8 × 104 | 0–200 | 3750 | 3200 | 0–200 |
B1 | Training of previous week’s data | 0–1.3 × 104 | 0–28 | 4000 | 2700 | 0–12 |
B2 | Prediction of subsequent week’s data | 0–1495 | 0–12 | 800 | 400 | 0–2 |
C1 | Training of previous consecutive weeks of particular day | 0–4.5 × 104 | 0–50 (but starts at 0–250) | 5500 | 3000 | 0–50 (but starts at 0–250) |
C2 | Prediction of subsequent consecutive weeks of particular day | 0–6.0 × 104 | 0–400 | 3000 | 4000 | 0–400 |