Software Reliability Estimation Using Cubic Splines Network Model

Abstract

The term quality in general, is a feeling. Thus, it is hard to describe consistently as a feeling is not consistent. Software quality is essentially a kind of quality particularly associating with software. Thus, the term software quality is also hard to describe. Hence, researchers use software quality models. Each software quality model consists of several factors which affect the software quality and they are called software quality factors. Software reliability is one of such software quality factors in nearly all the software quality models. Hence, software in order to be a high quality one, all the quality factors including software reliability has to be guaranteed. However, it is evident that software reliability is not guaranteed in almost all the commercial software development. This has been due to the lack of accuracy of the reliability estimation and the time taken to estimate the reliability in existing software reliability estimation models or software reliability growth models. Among the commonly used software reliability growth models, Non Homogeneous Poisson Model (NHPP model) shows more accuracy than the other models. However, in order to estimate the reliability, it requires more input data (i.e. a minimum of twenty five failure data). Thus, it takes considerable time. In this thesis, a novel software reliability growth model called Cubic Spline Network model (CSN model) has been introduced for improved accuracy with respect to the existing models. The proposed model requires relatively smaller number of past failure data as input and thus, this research will prove that it is more practical to use in the commercial software developments. Cubic splines network model has sensitivity of tuning for smaller or higher reliability estimation which has also not been introduced in the literature