Software Quality Estimation with Case-Based Reasoning

Abstract

Abstract The software quality team of a software project often strives to predict the operational quality of software modules prior to software deployment. A timely software quality prediction can be used for enacting any preventive actions so as to reduce software faults from occurring during system operations. This is especially important for high-assurance systems where software reliability is very critical. The two most commonly used models for software quality estimation are, software fault prediction and software quality classification. Generally, such models use software metrics as predictors of a software module's quality, which is either represented by the expected number of faults or a class membership to quality-based groups. This study presents a comprehensive methodology for building software quality estimation models with case-based reasoning ( cbr ), a computational intelligence technique that is suited for experience-based analysis. A cbr system is a practical option for software quality modeling, because it uses an organization's previous experience with its software development process to estimate the quality of a currently under-development software project. In the context of software metrics and quality data collected from a high-assurance software system, software fault prediction and software quality classification models are built. The former predicts the number of faults in software modules, while the latter predicts the class membership of the modules into the fault-prone and not fault-prone groups. This study presents in-depth details for the cbr models so as to facilitate a comprehensive understanding of the cbr technology as applied to software quality estimation.