![](bg8.png)
4 LESSONS LEARNED
This section presents the lessons that we learned dur-
ing this experimental research:
1. PLA designs with diffused features can influ-
ence on the obtained results, what impacts on the cor-
relation of the investigated metrics as happen in the
Experiment I.
2. In spite we cannot attest the type of correlation
between the metrics SSC and SVC is better to opti-
mize one of them at once because similarity and vari-
ability are two naturally opposite concepts.
3. If the architect wants to prioritize the optimiza-
tion of the similarity of a SPL, the metric SSC can be
selected as a objective to the search process.
4. If the architect wants to prioritize the optimiza-
tion of the variability of a SPL and the PLA design
does not contain compound components, does not
make sense to select the metric AV. In this case, it
is better to select the metric SVC as a objective to the
search process.
5. In spite we cannot attest if there is correlation
between the metrics SVC and AV , we observe that,
according to the definition of these two metrics, for
PLA designs without compound components, the in-
crease in the number of variable components leads to
higher values of both SVC and AV. Thus, it seems
sufficient select one of these two metrics as objective
to the search process.
These lessons represent an important contribution
as they help to build an initial body of knowledge on
correlations between the investigated metrics and on
their use in the context of PLA design optimization
using multi-objective algorithms by MOA4PLA. The
lessons also provide insights to plan further experi-
ments related to the similarity and variability metrics.
5 CONCLUDING REMARKS
In this paper, an experimental research was conducted
to investigate the possible correlation between metrics
related to similarity and variability of PLA design.
Three experiments were carried out with the follow-
ing pair of metrics (SSC, SVC ), (SVC, AV ) and (SSC,
AV) involving four PLA designs.
The empirical results are inconclusive. So, it
was not possible to characterize the possible corre-
lation between the metrics. However, we learned
some lessons about the use of these metrics in the
context of PLA design optimization by the approach
MOA4PLA.
Further experiments should be performed with
other PLA designs to: (i) corroborate the behavior
of SSC and SVC about PLAs with diffused features,
(ii) evaluate the impact of AV on PLAs that contain
compound components, and (iii) improve the body of
knowledge on correlations between the metrics SSC,
SVC and AV in the PLA design optimization context.
ACKNOWLEDGEMENTS
The authors thank CNPq for financial support.
REFERENCES
Basili, V. R., Selby, R. W., and Hutchens, D. H. (1986). Ex-
perimentation in software engineering. IEEE Trans.
Softw. Eng., 12(7):733–743.
Coello, C. A. C., Lamont, G., and van Veldhuizen, D.
(2007). Evolutionary Algorithms for Solving Multi-
Objective Problems. Springer, 2nd edition.
Colanzi, T. E. and Vergilio, S. R. (2016). A feature-driven
crossover operator for multi-objective and evolution-
ary optimization of product line architectures. Journal
of Systems and Software, pages 126–143.
Colanzi, T. E., Vergilio, S. R., Gimenes, I. M. S., and
Oizumi, W. N. (2014). A search-based approach for
software product line design. In Proc. of SPLC, vol-
ume 1, pages 237–241.
Contieri Jr, A. C., Correia, G. G., Colanzi, T. E.,
de S. Gimenes, I. M., OliveiraJr, E. A., Ferrari, S.,
Masiero, P. C., and Garcia, A. F. (2011). Extend-
ing UML components to develop software product-
line architectures: Lessons learned. In Proc. of ECSA,
pages 130–138.
Deb, K., Pratap, A., Agarwal, S., and Meyarivan, T. (2002).
A fast and elitist multiobjective genetic algorithm:
NSGA-II. IEEE Transactions on Evolutionary Com-
putation, 6(2):182 –197.
Delgado, Y., Colanzi, T. E., Vergilio, S. R., and San-
tos, M. (2017). An enhanced evaluation model for
search-based product line architecture design. In XX
Ibero-American Conference on Software Engineering
(CIbSE2017). Submitted.
Donegan, P. M. and Masiero, P. C. (2007). Design issues
in a component-based software product line. In Proc.
of Brazilian Symposium on Software Components, Ar-
chitectures and Reuse (SBCARS), pages 3–16.
F
´
ederle, E. L., Ferreira, T., Colanzi, T. E., and Vergilio,
S. R. (2015). OPLA-Tool: A support tool for search-
based product line architecture design. In Proceed-
ings of the 19th International Conference on Software
Product Line, SPLC 15, pages 370–373, USA. ACM.
Gomaa, H. (2011). Software modeling and design: UML,
use cases, patterns, and software architectures. Cam-
bridge University Press.
Guizzo, G., Colanzi, T. E., and Vergilio, S. R. (2014). A
Pattern-Driven Mutation Operator for Search-Based
ICEIS 2017 - 19th International Conference on Enterprise Information Systems
540