Title:
Predicting Quality Requirements Necessary for a Functional Requirement based on Machine Learning
Author(s):
Ken Tanaka, Haruhiko Kaiya, and Atsushi Ohnishi.
Source:
In The Seventh International Conference on Software Engineering
Advances (ICSEA 2012), pp. 540-547, Lisbon, 18-23 Nov. 2012.
Abstract:
In the early stage of the software development,
quality requirements should be explicitly specified as well as
functional requirements. Software architecture and/or design
decision should be largely reconsidered if some quality requirement
is overlooked in the early stage. We thus propose
a technique for predicting quality requirements necessary for
each functional requirement. A functional requirement is represented
with a semi-formal language called eXtended Japanese
Requirements Description Language (X-JRDL), which is based
on the case grammar. In our previous work, the results of
the prediction largely depended on human such as domain
experts and requirements analysts because prediction rules
were manually written by them. We thus introduce machine
learning to avoid this problem. To predict quality requirements
necessary for any kinds of functional requirements, training
data should be appropriately chosen. We choose the training
data so that we can predict necessary quality requirements
for all types of functional requirements. Since semantically
impartial data are suitable for such training data and one
of the cases called concept is semantically dominant in an
X-JRDL sentence, we choose the training data set in which
any of the concepts evenly occurs. Through the experiments,
we confirm our technique works well for predicting necessary
quality requirements.
Related Paper(s):