Statistical Study of Software Reliability Growth Models

Abstract

In recent years the use of the computer software has spread tremendously. Most of the machines running with computer software. A failure of a program to do its required job may result in a great damage.An error free- program remains the unattainable dream of the software manufacturer and the target of the consumer. The target of the software manufacturer is to find a reliable statistical model to estimate the remaining errors in the program before sending it to market. This thesis focus on analytical study of seven of the most popular Non Homogenous Poisson Process software reliability Growth Models. These statistical models are used to predict software faults. The study concentrates on the evaluation and comparison of the selected models using five published sets of software errors data. The models' parameters equations are derived using the likely hood method. The calculation of models are obtained using the computer package MATHCAT. Microsoft Excel is also used to draw the graphs and calculate the analysis results The research aim is to implement a simple procedure to evaluate and compare the software reliability growth models. Some complicated and difficult to use methods are available in the literature. The proposed procedure include graphs, run test, chi square test of hypothesis. Models stability and sensitivity analysis are included in the procedure. The research explained the use of the model's parameters in calculating some software reliability measures such as testing reliability and estimating the optimal release time.This will help the management to take the decision related to the right time when to stop testing and send the product to the market. The Analysis results showed that none of the fitted models can be addressed as the best model. The predictive ability of the model changed with the data set. To reach at valid conclusions the research recommended that more than one model should be fitted to predict the remaining faults in a project. The area of evaluating , validating and comparing the software reliability growth models still needs more researches by either improving the existing methods or developing new ones. Translation to Arabic language in this field of research is an essential need