Research Article
Machine Learning Approach for Software Reliability Growth
Modeling with Infinite Testing Effort Function
Subburaj Ramasamy and Indhurani Lakshmanan
School of Computing, SRM University, Kattankulathur, Tamil Nadu 603203, India
Correspondence should be addressed to Indhurani Lakshmanan; indhurani.a@gmail.com
Received 2 February 2017; Revised 22 June 2017; Accepted 27 June 2017; Published 26 July 2017
Academic Editor: Erik Cuevas
Copyright © 2017 Subburaj Ramasamy and Indhurani Lakshmanan. his is an open access article distributed under the Creative
Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the
original work is properly cited.
Reliability is one of the quantiiable sotware quality attributes. Sotware Reliability Growth Models (SRGMs) are used to assess
the reliability achieved at diferent times of testing. Traditional time-based SRGMs may not be accurate enough in all situations
where test efort varies with time. To overcome this lacuna, test efort was used instead of time in SRGMs. In the past, inite test
efort functions were proposed, which may not be realistic as, at ininite testing time, test efort will be ininite. Hence in this paper,
we propose an ininite test efort function in conjunction with a classical Nonhomogeneous Poisson Process (NHPP) model. We
use Artiicial Neural Network (ANN) for training the proposed model with sotware failure data. Here it is possible to get a large
set of weights for the same model to describe the past failure data equally well. We use machine learning approach to select the
appropriate set of weights for the model which will describe both the past and the future data well. We compare the performance
of the proposed model with existing model using practical sotware failure data sets. he proposed log-power TEF based SRGM
describes all types of failure data equally well and also improves the accuracy of parameter estimation more than existing TEF and
can be used for sotware release time determination as well.
1. Introduction
Early Sotware Reliability Growth Models (SRGMs) represent
the relationship between the time to failure and the cumula-
tive number of faults detected till then. Many such SRGMs
have been proposed as parametric [1–14] and nonparametric
[15–18] models since the year 1972 to estimate future failure
occurrence times and assess the reliability growth of sotware
systems during the testing phase. he traditional SRGMs are
based on the premise that the mean value function of the
model follows either exponential growth [1, 3] or S-shaped
growth [2, 11] or both [4–8].
Some SRGMs have been proposed with testing efort
function (TEF) [11–13], since the fault detection and cor-
rection depend on eforts consumed such as test cases
executed, man-days expended, computer utilization time,
and other resources consumed, rather than only testing time
or calendar time. he efort based SRGMs proposed in the
past use exponential, Rayleigh, logistic, or Weibull distri-
butions to specify testing efort function (TEF) to denote
efort consumption during testing [11–13]. Although these
functions seem to give good result and can well it in some
cases, there is a fallacy in assuming inite total test efort at
an ininite time. Xie and Zhao proposed a Nonhomogeneous
Poisson Process (NHPP) reliability growth model based on
log-power distribution which is a graphical model where
itting of the data or not can be visualized in a graph before
parameter estimation [19].
In this paper, we propose using log-power [19] distribu-
tion to describe TEF in Goel and Okumoto [1] SRGM to
provide an SRGM with ininite TEF. We use Artiicial Neural
Network (ANN) for parameter estimation and apply machine
learning technique to determine the most suitable weights
for the proposed model that will it the past and future data
equally well. We study and compare the goodness of it (GoF)
performance of the proposed model with a popular test efort
function based SRGM. We use ANN for parameter estima-
tion uniformly in all cases since ANN improves the parameter
estimation accuracy and gives better goodness of it rather
than traditional statistical parametric models [15–18].
Hindawi
Mathematical Problems in Engineering
Volume 2017, Article ID 8040346, 6 pages
https://doi.org/10.1155/2017/8040346