Measuring Software Reliability: A Trend Using Machine Learning Techniques - Complex System Modelling and Control Through Intelligent Soft Computations - page 794

this class imbalanced attribute, the data level or algorithm based measure both are
reported in corresponding literature (Seiert et al. 2009 ). The measures suggested are
emphasizing on dimension reduction issues. Improved dimension reduction meth-
ods are suggested for the class imbalanced problem by means of Partial Least
Squares (PLS) (Barker and Rayens 2003 ), Linear Discriminant Analysis (LDA)
(Xue and Titterington 2008 ) and Principle Component Analysis (PCA) (Jiang 2009 ;
Ma et al. 2012 ).
These postulates of software engineering motivate to deploy couple of machine
learning e.g. Support Vector Machine which could provide immediate advantage of
statistical estimation of reliability. As the proposed method has been gradually
shifted from statistical simulation to machine learning techniques, therefore certain
relevant observations have been recorded. At the beginning, the relative error
versus time graph obtained from SVR shows that a as more failure data from
previous weeks are collected during the software development and testing process,
learning is more ef

cient as a result of which parameters are estimated to values that
make the regression function better fit the given dataset. This fosters more accurate

prediction about the number of failures in succeeding week. The highlight of the
chapter is to motivate the inclusion of many other relevant machine learning
techniques. The remaining part of the chapter has been organized as follows: Sect. 2
lists out the different metrics those are signi

cant while presenting the analysis,
followed by the Sect. 3 with all relevant works on reliability measure. Section 4
concentrates on the data set concerning the importance for empirical and theoretical
validation of software reliability and Sect. 5 elaborates the inclusion of machine
learning and experimental analysis. Finally Sect. 6 gives the conclusion and dis-
cusses further scope of research.

2 Matrices to Measure Software Reliability: List
of Definitions

Conventionally, the concept of reliability in terms of failure data needs to be
properly measured by various means during software development and operational
phases. However, software failures are always design failures. Often the system
continues to be available in spite of the fact that a failure has occurred. Various
metrics used for measuring software reliability are described in Table 1 .

3 Relevant Recent Works: Software Reliability Measures

Various models of software reliability have been proposed, discussed, modi ed,

and formalized mainly since 1970s, although some of them have also suffered from
a lot of critiques. Software reliability analysis which was at the beginning (in
sixties) based on proof of correctness had passed to a period of stochastic modeling