IUP Publications Online
Home About IUP Magazines Journals Books Archives
     
Recommend    |    Subscriber Services    |    Feedback    |     Subscribe Online
 
The IUP Journal of Computer Sciences :
SPC-Based Software Reliability Using Modified Genetic Algorithm:
Inflection S-Shaped Model
:
:
:
:
:
:
:
:
:
 
 
 
 
 
 
 

In order to assess software reliability, many Software Reliability Growth Models (SRGMs) have been proposed in the past 40 years. In principle, two widely used methods for the parameter estimation of SRGMs are the Maximum Likelihood Estimation (MLE) and the Least Squares Estimation (LSE). However, the approach of these two estimations may impose some restrictions on SRGMs, such as the existence of derivatives from formulated models or the needs for complex calculation. In this paper, a Modified Genetic Algorithm (MGA) is proposed to assess the reliability of software considering the time domain software failure data and SPC using inflection S-shaped model which is Non-Homogenous Poisson Process (NHPP)-based. Experiments based on real software failure data are performed and the results show that the proposed Genetic Algorithm (GA) is more effective and faster than traditional algorithms.

 
 
 

Software reliability assessment is important to evaluate the quality of software system, since it is one of the most important attributes of software. One of the most difficult problems of software industry is to ship a reliable product. Therefore, it is necessary to have accurate and fast estimation techniques for verifying software reliability. For four decades, many Software Reliability Growth Models (SRGMs) have been proposed in estimating the reliability growth of software products. SRGMs can be used to depict the behavior of observed software failures characterized by either times of failures (i.e., time domain data) or by the number of failures at fixed times (i.e., interval domain data) (Lyu, 1996).

The parameters of SRGMs are generally unknown and have to be estimated based on the collected failure data. Two of the most popular estimation techniques are Maximum Likelihood Estimation (MLE) and Least Squares Estimation (LSE) (Ohba, 1984; and Goel, 1985). In fact, MLE and LSE involve the property of probability theory and statistical analysis. Thus, this may impose some restrictions on the parameter estimation of SRGMs (Minohara and Tohma, 1995; and Costa et al., 2007) such as the continuity, the unimodality, the existence of derivatives from formulated models and the complex likelihood function. The method of MLE estimation by solving a set of simultaneous equations is better in deriving confidence intervals. The method of LSE minimizes the sum of squares of the deviations between what we actually observe and what we expect. Nevertheless, LSE is suitable for fitting data from small to medium sample sizes (Wood, 1996), while MLE is considered to be a better statistical estimator for large sample sizes. In particular, when the formulated model of SRGMs is complicated or the sample size of failure data is large, these two estimation techniques may not be effective to find out the optimal solutions and generally require to be solved numerically. Hence, more effective and applicable approaches for the parameter estimation of SRGMs may be necessary.

 
 
 

Computer Sciences IUP Journal, Software reliability, Inflection S-shaped model, Time domain data, Mean value function, Modified Genetic Algorithm (MGA), Nonhomogenous Poisson Process (NHPP)