Table 5.
GEA | 7–10 | 4.5–7 | 1–5.5 | 2nd level | ||||||||
Threshold | 0.1 | 0.05 | 0.005 | 0.1 | 0.05 | 0.005 | 0.1 | 0.05 | 0.005 | 0.1 | 0.05 | 0.005 |
TP | 121 | 119 | 113 | 112 | 110 | 100 | 116 | 113 | 104 | 116 | 115 | 102 |
TN | 182 | 188 | 205 | 191 | 195 | 205 | 159 | 164 | 178 | 186 | 193 | 198 |
FP | 33 | 27 | 10 | 24 | 20 | 10 | 56 | 51 | 37 | 29 | 22 | 17 |
FN | 38 | 40 | 46 | 47 | 49 | 59 | 43 | 46 | 55 | 43 | 44 | 57 |
Precision | 0.79 | 0.82 | 0.92 | 0.82 | 0.85 | 0.91 | 0.67 | 0.69 | 0.74 | 0.80 | 0.84 | 0.86 |
Recall | 0.76 | 0.75 | 0.71 | 0.70 | 0.69 | 0.63 | 0.73 | 0.71 | 0.65 | 0.73 | 0.72 | 0.64 |
F1-Measure | 0.77 | 0.78 | 0.80 | 0.76 | 0.76 | 0.74 | 0.70 | 0.70 | 0.69 | 0.76 | 0.78 | 0.73 |
PRR | 7–10 | 4.5–7 | 1–5.5 | 2nd level | ||||||||
Threshold | 1 | 1.5 | 5 | 1 | 1.5 | 5 | 1 | 1.5 | 5 | 1 | 1.5 | 5 |
TP | 112 | 96 | 48 | 110 | 93 | 39 | 111 | 83 | 6 | 109 | 90 | 27 |
TN | 126 | 144 | 197 | 150 | 172 | 208 | 152 | 176 | 211 | 166 | 184 | 211 |
FP | 89 | 71 | 18 | 65 | 43 | 7 | 63 | 39 | 4 | 49 | 31 | 4 |
FN | 47 | 63 | 111 | 49 | 66 | 120 | 48 | 76 | 153 | 50 | 69 | 132 |
Precision | 0.56 | 0.57 | 0.73 | 0.63 | 0.68 | 0.85 | 0.64 | 0.68 | 0.60 | 0.69 | 0.74 | 0.87 |
Recall | 0.70 | 0.60 | 0.30 | 0.69 | 0.58 | 0.25 | 0.70 | 0.52 | 0.04 | 0.69 | 0.57 | 0.17 |
F1-Measure | 0.62 | 0.59 | 0.43 | 0.66 | 0.63 | 0.38 | 0.67 | 0.59 | 0.07 | 0.69 | 0.64 | 0.28 |
Performance with GEA (top) and PRR (bottom) using different abstraction levels and thresholds. Bold numbers indicate best performance throughout all configurations (individually for GEA and PRR-based configurations)