Table 2 Four-category accuracies: UE-based raw classification vs aggregation decision
From: Anomaly detection in multi-tiered cellular networks using LSTM and 1D CNN
Accuracies (%) | ||||||
|---|---|---|---|---|---|---|
2 dB | 5 dB | 8 dB | ||||
UE-based classification | Ensemble averaging | UE-based classification | Ensemble averaging | UE-based classification | Ensemble averaging | |
LSTM | ||||||
30m | 65.75 | 79.08 | 51.40 | 59.18 | 43.50 | 50.83 |
40m | 60.43 | 77.92 | 44.17 | 55.83 | 40.17 | 50.51 |
50m | 55.70 | 75.83 | 41.04 | 52.50 | 39.83 | 48.75 |
1D CNN | ||||||
30m | 65.75 | 80.10 | 48.60 | 56.38 | 42.13 | 46.43 |
40m | 59.83 | 74.58 | 43.45 | 55.00 | 40.28 | 46.25 |
50m | 54.77 | 73.33 | 41.67 | 52.50 | 36.54 | 46.25 |