@MastersThesis{Balera:2017:AlGeCa,
author = "Balera, Juliana Marino",
title = "Um algoritmo para gera{\c{c}}{\~a}o de casos de teste
combinatorial via matriz de cobertura com n{\'{\i}}veis
variados",
school = "Instituto Nacional de Pesquisas Espaciais (INPE)",
year = "2017",
address = "S{\~a}o Jos{\'e} dos Campos",
month = "2017-02-14",
keywords = "teste de software, T-Tuple Reallocation, designs combinatoriais,
matriz de cobertura com n{\'{\i}}veis variados, experimento
controlado, software testing, T-Tuple Reallocation, combinatorial
designs, mixed-level covering arrays, controlled experiment.",
abstract = "Na perspectiva de sistemas complexos, como softwares desenvolvidos
para aplica{\c{c}}{\~o}es espaciais tais como sat{\'e}lites,
bal{\~o}es estratosf{\'e}ricos e foguetes, existem sempre riscos
relacionados ao mau funcionamento do produto que podem causar
danos ao meio ambiente, grandes perdas financeiras, ou o pior,
perda de vidas. Para minimizar ao m{\'a}ximo esses riscos, {\'e}
necess{\'a}rio que o processo de teste desses sistemas ocorra de
forma rigorosa e eficiente. Como n{\~a}o {\'e} poss{\'{\i}}vel
testar tais produtos exaustivamente, dada a larga gama de casos de
teste poss{\'{\i}}veis, {\'e} fundamental, portanto, que se
tenham dispon{\'{\i}}veis m{\'e}todos para a
gera{\c{c}}{\~a}o/sele{\c{c}}{\~a}o de casos de teste que
possuem grande potencial de revela{\c{c}}{\~a}o de defeitos, e
que possuam custo reduzido. Nessa dire{\c{c}}{\~a}o, designs
combinatoriais v{\^e}m chamando aten{\c{c}}{\~a}o da comunidade
de teste de software para gerar conjuntos de casos de testes
menores (menor custo para executar) e eficientes (capacidade de
encontrar defeitos no software). Diante disso, essa
disserta{\c{c}}{\~a}o de mestrado tem como objetivo apresentar
uma nova forma de gerar conjuntos de casos de teste via designs
combinatoriais, sendo que tais casos de teste tenham custo menor e
efici{\^e}ncia compar{\'a}vel {\`a} solu{\c{c}}{\~o}es
j{\'a} existentes na literatura. Ent{\~a}o, um algoritmo,
denominado T-Tuple Reallocation (TTR; Realoca{\c{c}}{\~a}o de
T-Tuplas), para gerar casos de teste de software via designs
combinatoriais, especificamente via a t{\'e}cnica de Matriz de
Cobertura com N{\'{\i}}veis Variados (MCNV), foi desenvolvido. A
ideia geral do TTR {\'e} derivar uma MCNV M por meio da
cria{\c{c}}{\~a}o e realoca{\c{c}}{\~a}o de t-tuplas para a
matriz M, considerando um par{\^a}metro chamado meta
(\$\zeta\$). Dois experimentos controlados e um
quasiexperimento foram realizados para comparar o TTR com outros
quatro algoritmos/ferramentas bastante conhecidos que geram MCNVs.
No primeiro experimento controlado, comparou-se duas perspectivas
de custo considerando a vers{\~a}o 1.1 do algoritmo TTR: tamanho
das suites de teste e tempo para gerar as suites de teste.
Al{\'e}m disso, realizou-se uma an{\'a}lise de similaridade
entre esses conjuntos. No segundo experimento controlado, foi
considerada uma vers{\~a}o melhorada do algoritmo TTR,
vers{\~a}o 1.2, e comparou-se com os mesmos quatro
algoritmos/ferramentas anteriores, mas considerando somente a
perspectiva de custo relacionada ao tamanho das suites de teste e
an{\'a}lise de similaridade. Por fim, um quasiexperimento foi
realizado onde comparou-se a efici{\^e}ncia entre o TTR 1.2 e as
outras quatro solu{\c{c}}{\~o}es, usando an{\'a}lise de
mutantes e aplicando a um estudo de caso da {\'a}rea espacial. As
conclus{\~o}es dessas tr{\^e}s avalia{\c{c}}{\~o}es rigorosas
s{\~a}o que o TTR foi o algoritmo que apresentou melhor custo
(menor quantidade de casos de teste para serem executados), mas
que n{\~a}o h{\'a} diferen{\c{c}}a de efici{\^e}ncia entre o
TTR e as demais solu{\c{c}}{\~o}es. Al{\'e}m disso, as suites
de teste n{\~a}o s{\~a}o similares, comparando o TTR com as
outras solu{\c{c}}{\~o}es. Desse modo, pode-se afirmar que o TTR
foi superior aos demais algoritmos/ferramentas pois teve mesma
efici{\^e}ncia mas melhor custo. ABSTRACT: With respect to
complex systems, such as software developed for space applications
like satellites, stratospheric balloons and rockets, there are
always risks related to product malfunctioning that can cause
damage to the environment, great financial losses, or worse, loss
of lives. To minimize these risks, the testing process of these
systems must be rigorous and efficient. Since it is not possible
to test such products exhaustively, given the wide range of
possible test cases, it is critical, therefore, that there are
available methods for the generation/selection of test cases which
have great potential for defects detection and a reduced cost. In
this direction, combinatorial designs have drawn attention of the
software testing community to generate sets of smaller (lower cost
to run) and efficient (ability to find software defects) test
cases. Therefore, this master dissertation aims to present a new
way for generating sets of test cases via combinatorial designs,
where such test cases have a lower cost and efficiency comparable
to solutions already existing in the literature. Then, an
algorithm, called T-Tuple Reallocation (TTR), to generate software
test cases via combinatorial designs, specifically via the
Mixed-Level Covering Array technique (MCA) was developed. The main
reasoning behind TTR is to derive an MCA M by creating and
relocating t-tuples to the matrix M, considering a parameter
called goal (\$\zeta\$). Two controlled experiments and one
quasiexperiment were performed to compare TTR with four other
well-known algorithms/tools that generate MCAs. In the first
controlled experiment, version 1.1 of TTR was compared considering
two cost perspectives: size of the test suites and time to
generate the test suites. In addition, a similarity analysis was
accomplished between these sets. In the second controlled
experiment, an improved version of TTR, version 1.2, was compared
with the same four previous algorithms/tools, but only in the cost
perspective related to the size of the test suites and similarity
analysis. Finally, a quasiexperiment aimed to assess the
efficiency between TTR 1.2 and the other four solutions was
carried out, via mutation testing and a space application case
study. The conclusions of these three rigorous evaluations are
that TTR was the algorithm that presented the better cost (smaller
number of test cases to execute), but that there is no difference
in efficiency between TTR and the other solutions. In addition,
the test suites are not similar, comparing TTR with the other
approaches. Thus, it can be asserted that the TTR algorithm was
superior to the other algorithms/tools because it had the same
efficiency but better cost.",
committee = "Mendes, Celso Luiz (presidente) and Santiago J{\'u}nior,
Valdivino Alexandre de (orientador) and Guerra, Eduardo Martins
and Martins, Luiz Eduardo Galv{\~a}o",
englishtitle = "An algorithm for the generation of combinatorial test cases via
mixed-level covering arrays",
language = "pt",
pages = "109",
ibi = "8JMKD3MGP3W34P/3NK7TBE",
url = "http://urlib.net/ibi/8JMKD3MGP3W34P/3NK7TBE",
targetfile = "publicacao.pdf",
urlaccessdate = "27 abr. 2024"
}