Test effort optimization by prediction and ranking of fault-prone software modules

Abstract

Identification of fault-prone or not fault-prone modules is very essential to improve the reliability and quality of a software system. Once modules are categorized as fault-prone or not fault-prone, test effort are allocated accordingly. Testing effort and efficiency are primary concern and can be optimized by prediction and ranking of fault-prone modules. This paper discusses a new model for prediction and ranking of fault-prone software modules for test effort optimization. Model utilizes the classification capability of data mining techniques and knowledge stored in software metrics to classify the software module as fault-prone or not fault-prone. A decision tree is constructed using ID3 algorithm for the existing project data. Rules are derived form the decision tree and integrated with fuzzy inference system to classify the modules as either fault-prone or not fault-prone for the target data. The model is also able to rank the fault-prone module on the basis of its degree of fault-proneness. The model accuracy are validated and compared with some other models by using the NASA projects data set of PROMOSE repository.