Search Based Software Engineering: Techniques, Taxonomy, Tutorial - Empirical Software Engineering and Verification - page 46

empirical study, by selecting some non trivial real world examples to which the
new SBSE approach can be applied. The most natural questions to address are
those of eciency and effectiveness. That is, the study should, at least, answer
the questions: how good are the solutions you find and how much computational
(and/or human) effort is required to obtain them?
Step 7: One of the attractive aspects of SBSE is the way it re-unites different
areas of software engineering, as explained in Section 2. A good paper on SBSE
will contain a thorough account of the related work and this may help to achieve
this re-unification goal, drawing together apparently unrelated areas of software
engineering. In so doing, SBSE work can play a vital role in the development
of the wider field of Software Engineering itself, providing a more solid search-
based understanding of the underlying optimisation problems that are found in
each application area.
Using SBSE, we seek to apply search algorithms to software engineering prob-
lems so there are two natural sources of related work for any paper on SBSE;
the previous work that tries to solve the same (or similar) Software Engineering
problem(s) and the previous work that uses a similar search based approach. You
may find that two apparently quite different software engineering problems have
been attacked using the same (or similar) search based formulation (perhaps
representation is shared or a similar fitness can be used).
An SBSE paper can be considerably enhanced by exploring these links, since
such connections mean that the paper canhaveanimpactonatleasttwosoft-
ware engineering domains, rather than merely the one for which the results are
presented. Hopefully, in developing a related work section, the taxonomy in Sec-
tion 9 will be helpful. The many surveys on SBSE are also a source of valuable
summary information concerning potentially related techniques and application
areas.
Step 8: At this point, the researcher has sucient information and results to
consider writing a paper. Naturally, there will be a choice about where to send
the paper that can only be made by the author(s). There is also the question
of how to set the problem formulation, research questions and results into as
format that will appeal to (firstly) referees and (ultimately) readers.
There are many excellent papers that give guidance on how to write good
software engineering papers, such as that by Mary Shaw [86]. Those papers that
present results on SBSE generally (though not exclusively) fall in the category
of empirical software engineering papers, for which the systematic review of Ali
et al. [4] sets out useful guidelines relevant to SBSE.
11 SBSE Limitations and Techniques for Overcoming
Them
In this final section we review some of the problems and issues that can arise
when using SBSE and some simple techniques for overcoming them. Many of
these issues are common to all approaches to optimisation based on algorithms