ABSTRACT

208Software quality analysis has become a key factor of analysis in development due to the considerable increase in the use of such systems in the current world. Owing to this, several models for describing this quality in terms of reliability have been proposed, some of which are presented in this study. The purpose of this study is to evaluate the performance of several statistical software reliability models given the experimental times between failures obtained from Musa's software reliability data benchmark. For the analytics, an R-based platform was implemented with parameter estimation routines for a variety of statistical software reliability structures. This tool provided a visual representation of the behavior of the number of failures described by each model. The statistical models provided an effective tool along with a theoretical background to describe software reliability. Owing to the complexity of the functions, parameter estimation may pose fairly complicated challenges from the computational point of view. The sort of models described in this work could be used to characterize software quality from the perspective of its reliability given the collected information regarding times between failures. These tools could also be integrated to an online system which creates an information loop between failure detection, reliability measurement, and debugging processes. The application developed in Shiny by RStudio has been implemented for monitoring in the virtual branch of the largest financial institution in Colombia. This research provides a starting point for developing a software reliability analysis tool, which could be expanded to a greater number of models and to further techniques related to the study of software reliability. These tools not only include the parameter estimation for such models but also a visualization of the estimations and predictions made by each model, providing valuable insights on their behavior.