Search Based Software Engineering: Techniques, Taxonomy, Tutorial - Empirical Software Engineering and Verification - page 35

ii. to highlight to future research directions by identifying unexplored opportu-
nities.
iii. to allow the development of automated search tools to enable effective and
ecient search of SBSE problems in any particular repository.
iv. to facilitate re-use, especially regarding approximation algorithms and the-
oretical bounds.

The proposed taxonomy of Search Based Software Engineering problems will
involve Perspectives, Dimensions and Characteristics.
The two Perspectives, SOFTWARE ENGINEERING and OPTIMISATION,
will reflect the different points of view under which a particular SBSE problem
can be analysed. The Dimensions, as in other taxonomies, will represent, for
each Perspective, the SBSE problem features. Finally, for each Dimension, the
Characteristics will correspond to the possible feature values under which a par-
ticular SBSE problem can be identified. For all Dimensions, the Characteristics
are collectively exhaustive. However, only the Dimensions “Objective Space Di-
mensionality”, “Instance Space Characterisation”, “Constrained” and “Problem
Linearity” are mutually exclusive, for all others, more than one Characteristic
may be selected for a particular SBSE problem.
In Tables 2 and 3, the proposed taxonomy of Search Based Software Engi-
neering problems is presented.
For the SOFTWARE ENGINEERING Perspective (Table 2), four Dimen-
sions are identified: “Software Development Stage(s)”, “Software Development
Model(s)”,

“Main Subject Descriptor(s)” and “Main Implicit Subject

Descriptor(s)”.

1. The “Software Development Stage(s)” positions the problem under one,
or more, stages in the software engineering process. The Characteristics avail-
able for this Dimension are representative of the standard software develop-
ment process.
2. Next, the “Software Development Model(s)” identifies a particular set
of development models in which the problem occurs.
3. The Dimension named “Main Subject Descriptor(s)” describes the soft-
ware engineering subject addressed by the problem. The Characteristics
present in this Dimension were obtained from the 1998 version of the ACM
Computing Classification System [1]. More specifically, the possible values
for this feature are those defined as a “Subject Descriptor” under the level
D.2 (SOFTWARE ENGINEERING), in the third level of the classification
structure, with values, and corresponding subjects, ranging from D.2.0 (Gen-
eral) to D.2.13 (Reusable Software) and D.2.m (Miscellaneous).
4. Finally, the “Main Implicit Subject Descriptor(s)” Dimension details
the previous subject descriptor(s), by allowing the selection of the more spe-
cific subject descriptors present in the fourth level of the ACM Computing
Classification System [1], once again under the level D.2 (SOFTWARE EN-
GINEERING).