Global Sensitivity Analysis of Predictor Models in Software Engineering

Predictor models are an important tool in software projects for quality and cost control as well as management. There are various models available that can help the software engineer in decision-making. However, such models are often difficult to apply in practice because of the amount of data needed. Sensitivity analysis offers provides means to rank the input factors w.r.t. their importance and thereby reduce and optimise the measurement effort necessary. This paper presents an example application of global sensitivity analysis on a software reliability model used in practice. It describes the approach and the possibilities offered.