Table 2 Mean overall cost improvement of A-greedy* over A-greedy (%)
From: Incorporating change detection in the monitoring phase of adaptive query processing
Experiment 1 | |||
Abrupt 1000 K | Abrupt 200 K | Abrupt 50 K | |
A-greedy*/ADWIN2 | 11.58 % | 10.99 % | 12.08 % |
A-greedy*/Martingale Test | 11.70 % | 12.12 % | 14.08 % |
A-greedy*/ChangeFinder | 11.56 % | 11.63 % | 12.19 % |
A-greedy*/Meta-algorithm | 8.64 % | 4.79 % | 6.41 % |
A-greedy*/ β-CUSUM | 14.06 % | 13.55 % | 14.89 % |
Experiment 2 | |||
Abrupt 1000 K | Abrupt 200 K | Abrupt 50 K | |
A-greedy*/ADWIN2 | 36.92 % | 27.49 % | 20.75 % |
A-greedy*/Martingale Test | 31.76 % | 29.83 % | 31.16 % |
A-greedy*/ChangeFinder | 31.83 % | 30.48 % | 31.06 % |
A-greedy*/Meta-algorithm | 31.19 % | 27.68 % | 30.76 % |
A-greedy*/ β-CUSUM | 36.58 % | 34.54 % | 34.58 % |
Experiment 3 | |||
Abrupt 1000 K | Abrupt 200 K | Abrupt 50 K | |
A-greedy*/ADWIN2 | 6.04 % | 6.79 % | 9.80 % |
A-greedy*/Martingale Test | 6.57 % | 7.62 % | 9.60 % |
A-greedy*/ChangeFinder | 6.39 % | 6.83 % | 7.25 % |
A-greedy*/Meta-algorithm | 2.89 % | -1.02 % | 0.002 % |
A-greedy*/ β-CUSUM | 8.31 % | 8.62 % | 9.73 % |