Skip to main content

Table 3 % Mean overall cost improvement of A-greedy* over A-greedy

From: Incorporating change detection in the monitoring phase of adaptive query processing

Experiment 4

 

Transition

Transition

Transition

 

length 10 K

length 25 K

length 50 K

A-greedy*/ADWIN2

19.50 %

20.38 %

21.07 %

A-greedy*/Martingale Test

20.65 %

20.82 %

21.35 %

A-greedy*/ChangeFinder

20.25 %

20.77 %

21.06 %

A-greedy*/Meta-algorithm

13.54 %

14.73 %

17.99 %

A-greedy*/ β-CUSUM

23.90 %

23.43 %

23.53 %

Experiment 5

 

Transition

Transition

Transition

 

length 10 K

length 25 K

length 50 K

A-greedy*/ADWIN2

11.62 %

10.98 %

10.46 %

A-greedy*/Martingale Test

14.62 %

14.26 %

16.19 %

A-greedy*/ChangeFinder

11.94 %

13.15 %

15.78 %

A-greedy*/Meta-algorithm

-2.52 %

1.88 %

0.97 %

A-greedy*/ β-CUSUM

16.71 %

16.35 %

17.75 %

Experiment 6

 

Transition

Transition

Transition

 

length 10 K

length 25 K

length 50 K

A-greedy*/ADWIN2

27.69 %

28.53 %

33.56 %

A-greedy*/Martingale Test

27.93 %

26.30 %

32.02 %

A-greedy*/ChangeFinder

28.19 %

27.49 %

31.56 %

A-greedy*/Meta-algorithm

22.92 %

23.00 %

30.56 %

A-greedy*/ β-CUSUM

31.70 %

32.15 %

36.23 %

Experiment 7

 

Transition

Transition

Transition

 

length 10 K

length 25 K

length 50 K

A-greedy*/ADWIN2

17.82 %

18.70 %

18.32 %

A-greedy*/Martingale Test

19.15 %

19.69 %

19.00 %

A-greedy*/ChangeFinder

18.63 %

19.38 %

18.74 %

A-greedy*/Meta-algorithm

11.62 %

11.81 %

15.22 %

A-greedy*/ β-CUSUM

21.10 %

21.62 %

20.73 %

  1. The winning cases are shown in bold