Database Reference
In-Depth Information
Table 5. Specificity evaluation
k
=60
k
=50
k
=40
k
=30
k
=20
k
=10
k
VOMM
price eturn
0.896
0.918
0.941
0.963
0.983
0.998
_
k
VOMM
price ange
0.894
0.918
0.941
0.965
0.988
0.998
_
k
VOMM
trade
0.889
0.913
0.933
0.958
0.983
0.998
_
amounts
V-BOMM
k
0.921
0.938
0.963
0.975
0.993
1.00
P-BOMM
k
0.903
0.926
0.950
0.975
0.990
1.00
Table 6. Precision evaluation
k
=60
k
=50
k
=40
k
=30
k
=20
k
=10
k
VOMM
price eturn
0.300
0.340
0.400
0.500
0.650
0.900
_
k
VOMM
price ange
0.283
0.340
0.400
0.533
0.750
0.900
_
k
VOMM
trade
0.250
0.300
0.325
0.433
0.650
0.900
_
amounts
V-BOMM
k
0.385
0.444
0.531
0.630
0.842
1.000
P-BOMM
k
0.350
0.400
0.500
0.667
0.800
1.000
Table 7. Recall evaluation
k
=60
k
=50
k
=40
k
=30
k
=20
k
=10
k
VOMM
price eturn
_
0.857
0.810
0.762
0.7143
0.619
0.429
k
VOMM
price ange
_
0.810
0.810
0.762
0.762
0.714
0.429
k
VOMM
trade
_
amounts
0.714
0.714
0.619
0.619
0.619
0.429
V-BOMM
k
0.952
0.952
0.810
0.810
0.762
0.429
P-BOMM
k
1.000
0.952
0.952
0.952
0.762
0.477
see that the accuracy, precision and specificity
decrease with the increase of the
k
value. How-
ever, the recall increases with the increase of
k
.
By comparing the methods, it is obvious that our
proposed V-BOMM and P-BOMM perform better
than the outlier mining on single outlier mining,
and they have improved accuracy, precision, recall
and specificity.
Search WWH ::
Custom Search