Information Technology Reference
In-Depth Information
1 Introduction
The trend of growth of the size and complexity of software systems shows an
exponential growth and will continue in the future. With the increase in require-
ments and demands for, and dependencies on computers, the probability of failures
in software also increase which may cause serious, even fatal consequences to
various systems, especially time-critical systems. As a result, decent quality of
information and support systems has become a major concern for our society.
Reliability is the most important aspect of quality, and software reliability is de
ned
as the probability of failure-free software operation for a speci
ed period of time in
a speci
ed environment (Musa
1973
; Musa and Okumoto
1973
; Huang and Lyu
2011
). Hence, software reliability engineering (SRE) has become an interesting
topic of research and is developing very rapidly. Reliability assessment methods
and improvement techniques are of signi
cance to software managers, users as well
as practitioners.
In this chapter, an effort has been initiated to measure reliability of software
application while incorporating machine learning techniques. The motivation of
such work is primarily bi-focal: Firstly, conventional metric measures and statistical
estimation could be questioned in terms of ef
cacy and moreover machine learning
could be a suitable choice; hence search based software engineering could yield an
emerging vertical as reliability measure. Primarily it will comprise of broad
activities in terms measuring the reliability of the software product or application
(Gupta et al.
2011
).
Estimation: Determination of current software reliability by applying statistical
inference techniques to failure obtained during system test or system operation.
It is measure regarding achieved reliability from past until current point.
Prediction: Determination of future software reliability based on available
software metrics and measures. There are two possible cases:
Failure data is available: This is applicable to testing and operation stage
wherein estimation techniques can be used to parameterize and verify soft-
ware reliability models to perform future reliability prediction.
-
Failure data not available: This is applicable to design and coding stage
where metrics obtained from software development process and character-
istics of developed product can be used to determine reliability of software
upon testing or delivery.
-
Considering the growth model of reliability analysis (Gholizadeh et al.
2012
) the
recent reliability artifacts concentrate several verticals of reliability including the
fuzzy Bayesian system reliability assessment also has been initiated. The process is
based on prior two-parameter exponential distribution for estimating fuzzy map of
reliability under squared error for any given soft-ware domain under test. Although
software reliability growth model (SRGM) basically predicts the fault detection
coverage in software testing phases. The general problem that is encountered is to
minimize the number of remaining faults for a given
fixed amount of testing effort
Search WWH ::
Custom Search