Methods to Predict the Number of Software Faults Using Weibull Distribution
Koji Tanaka, Kazuhiko Tsuda · 2016
Quality improvement and control of software is an important agenda for software development firms and, thus, many companies has been conducting measurement and prediction of quality using metrics. As a method to predict the quality based on a history of removed faults of relevant project, there is the Rayleigh model. This model uses the fact that the rate of removed faults per unit of time generally follows the distribution of Rayleigh. However, in the Rayleigh model, the rate of removed faults of some projects do not follow the Rayleigh distribution, thus the gap between the predicted value and performance value becomes significant. Therefore, this study aims to develop a method of predicting numbers of faults that can also predict the quality, by using the quality standard value which shows the Weibull distribution and the knowledge level for software development. Additionally, the evaluation result shows that this method has better prediction accuracy than the Rayleigh model.