A software reliability model with time-dependent fault detection and fault removal

Abstract

The common assumption for most existingsoftware reliability growth models is that fault is independent and can be removed perfectly upon detection. However, it is often not true due to various factors including software complexity, programmer proficiency, organization hierarchy, etc. In this paper, we develop a software reliability model with considerations of fault-dependent detection, imperfect fault removal and the maximum number of faults software. The genetic algorithm (GA) method is applied to estimate the model parameters. Four goodness-of-fit criteria, such as mean-squared error, predictive-ratio risk, predictive power, and Akaike information criterion, are used to compare the proposed model and several existing software reliability models. Three datasets collected in industries are used to demonstrate the better fit of the proposed model than other existing software reliability models based on the studied criteria.