Using ANN to Study the Reliability Growth of Multiple Versions of Software Systems

The assessment of software’s reliability via developed mathematical models is one of those areas which is studied a lot by management. With new innovations in science and industry, the applicability of artificial neural networks, i.e., ANN has been employed to assess software’s reliability paradigm. ANN is a Machine Learning algorithm which processes the information/data, as it moves through a series of interconnected neurons, which constitute a neural structure. It was observed that, with a perfect combination of assigned weights and activation function, ANN can predict fault-occurrence pattern of a software as an output at its output node. Keeping this in mind, the authors have made some alterations in its network architecture, and have developed a model which resembles the fault occurrence phenomenon of a software system having different severity of faults. Further, this concept has been extended to assess the reliability indices of multi-release of software system. The performance analysis of developed structure is showcased with the help of comparison criterions, on which it was performing better than the traditionally-acknowledged procedures of parameter estimation.