Information Technology Reference
In-Depth Information
Table 3.
Data on accuracy with viewpoint threshold = 10
VP Threshold = 10
Light Control
Library
Manual
EA-Miner
Manual
EA-Miner
Precision
VP
6/9 = 66.7%
2/
10
= 20%
7/11 = 63.6%
5/
10
= 50%
Recall
VP
6/11 = 54.5%
2/11 = 18.2%
7/8 = 87.5%
5/8 = 62.5%
Precision
EA
6/8 = 75%
6/6 = 100%
7/10 = 70%
4/4 = 100%
Recall
EA
6/7 = 85.7%
6/7 = 85.7%
7/8 = 87.5%
4/8 = 50%
Regarding the precision and recall on early aspects identified, EA-Miner provided
a slight advantage over the manual approach. This is justified by the fact that the
tool's early aspect identification does not list many candidates and the ones listed are
already based on a more efficient knowledge (lexicon + semantic tag). In some cases
such as for example the recall of EA for library system the number can be low
because the lexicon was not properly populated for that specific domain. As explained
in Sect. 4 EA-Miner offers capabilities to update the lexicon minimizing this problem.
Table 4.
Data on accuracy with viewpoint threshold = 20
VP Threshold = 20
Light Control
Library
Manual
EA-Miner
Manual
EA-Miner
Precision
VP
6/9 = 66.7%
7/
20
= 35%
7/11 = 63.6%
6/
20
= 30%
Recall
VP
6/11 = 54.5%
7/11 = 63.6%
7/8 = 87.5%
6/8 = 75%
Precision
EA
6/8 = 75%
6/6 = 100%
7/10 = 70%
4/4 = 100%
Recall
EA
6/7 = 85.7%
6/7 = 85.7%
7/8 = 87.5%
4/8 = 50%
Table 5 shows data on the identified viewpoints by EA-Miner combined with the
use of guidelines and the experience of the requirements engineer. In this case the
precision and recall of viewpoints grows respectively to 90 and 81.8% for the light
control and to 88.8 and 100% in the library system. We can see that in this case the
accuracy of EA-Miner for viewpoint identification is much superior to the manual
approach. As we mentioned before the time spent in identification for this was not
exactly recorded but was still inferior than the manual approach (between 15-20 min
for light control and 25-30 to library).
Search WWH ::
Custom Search