Recent Advances of Exception Mining in Stock Market - Evolving Application Domains of Data Warehousing and Mining - page 225

Table 3. Output of experiments.

k =60 k =50 k =40 k =30 k =20 k =10

k

VOMM trade

18 17 16 15 13 9

_ amounts

k

VOMM trade

17 17 16 16 15 9

_ amounts

k

VOMM trade

15 15 13 13 13 9

_ amounts

V-BOMM k 20/52 20/45 17/32 17/27 16/19 9/9

P-BOMM k 21 20 20 20 16 10

outliers. The outcomes are vectors named
as P-BOMM k ( k =60,50,40,30,20,10).

pare the performance of the five methods according
to the four measures: accuracy, precision, recall
and specificity. They are calculated based on the
following formula:
Accuracy = ( TP + TN ) / ( TP + FN + FP + TN ) (5)
Specificity = TN / ( FP + TN ) (6)
Precision = TP / ( TP + FP ) (7)
Recall = TP / ( TP + FN ) (8)
The TP represents the number of detected
outliers those are real alerts; FP stands for the
number of detected which are not actual alerts;
FN represents the number of the identified nor-
mal days which are real alerts and TN stands for
the number of identified normal days which are
not alerts.
The Tables 4, 5, 6, & 7 show the values of
the accuracy, specificity, precision and recall
for different methods. From these tables, we can

The experimental results are shown in Table 3.
The columns stand for the factor k , the expected
number of outliers. For example, k =20 means that
the top 20 samples are regarded as outliers, while
the rest of the samples are regarded as normal.
The observation in each row stands for the alerts
which are exactly identified by corresponding
methods. For example, the value 16 on row 1
and column 4 means that 16 alerts are identified
by VOMM methods on price return time series.
One special case is the observation of V-BOMM,
where the left value stands for the number of real
alerts detected, while the right value stands for
the calculated number of outliers.
We evaluate the experiments results and com-

Table 4. Accuracy evaluation

k =60 k =50 k =40 k =30 k =20 k =10

k

VOMM price eturn

0.894 0.913 0.932 0.951 0.965 0.969

_

k

VOMM price ange

0.889 0.913 0.932 0.955 0.974 0.969

_

k

VOMM trade

0.88 0.904 0.918 0.941 0.965 0.969

_ amounts

V-BOMM k 0.922 0.939 0.955 0.967 0.981 0.972

P-BOMM k 0.908 0.927 0.951 0.974 0.979 0.974