Research Article
Recurrent Adaptive Classifier Ensemble for Handling Recurring Concept Drifts
Table 2
Prediction accuracies and average ranks of the five algorithms.
| | Dataset | RACE | CALMID | DUE | SOFEnsemble | AWDOB |
| | Hyperplane | 81.67 (1) | 69.85 (3) | 60.74 (4) | 76.61 (2) | 61.42 (5) | | Stagger | 79.34 (1) | 74.39 (2) | 64.62 (4) | 73.56 (3) | 63.52 (5) | | LED | 83.46 (1) | 70.35 (3) | 69.39 (4) | 79.62 (2) | 66.47 (5) | | SEA | 76.34 (2) | 73.49 (3) | 64.43 (5) | 80.28 (1) | 66.78 (4) | | Random Tree | 84.4 (1) | 68.76 (5) | 71.53 (3) | 81.62 (2) | 71.38 (4) | | Airlines | 78.59 (2) | 71.46 (3) | 61.34 (5) | 86.73 (1) | 63.32 (4) | | KDD99 | 69.28 (3) | 67.54 (4) | 81.36 (2) | 63.48 (5) | 83.12 (1) | | Covertype | 81.56 (1) | 73.37 (3) | 66.87 (4) | 81.36 (2) | 64.39 (5) | | Poker Hand | 84.31 (1) | 70.48 (3) | 69.38 (4) | 66.49 (5) | 82.34 (2) | | Sensor Data | 80.34 (1) | 72.57 (3) | 70.43 (4) | 79.67 (2) | 69.36 (5) | | Average ranks | 1.4 | 3.2 | 3.9 | 2.5 | 4.0 |
|
|
