Empirical Assessment of the Application of Machine Learning Techniques for Software Defect Prediction

Abstract

newline The cost of deleting a software bug increases ten times as it is floated onto the next phase of software development lifecycle (SDLC). This makes the task of the project managers difficult and also degrades the quality of the output software product. Therefore, software industry has stipulated the need for good quality software projects to be delivered on time and within budget. Software defect prediction (SDP) was proposed as a solution to the problem which could anticipate the defective modules and hence, deal with them in an efficient and effective manner in advance. It has permitted the project managers to anticipate the probable defects in the future and hence, give sufficient attention to such project components. SDP has led to the application of machine learning algorithms for building defect classification models using software metrics and defect proneness as the independent and dependent variables, respectively. SDP has emerged as a promising field since a last few years and very less work has been done in this significant area. Various machine learning methods are proposed and successfully applied in the literature. Several machine learning methods are available and there is a need to compare the performance of different machine learning methods as they give different results. Hence, more empirical studies which can be verified through replication are needed. Abstract ii Public data sets help in generalizing the results of the experiments performed earlier because they can be used to verify or refute the results obtained by other researchers. The use of public data sets in conducting empirical studies has also increased substantially. The above issues are addressed through different studies in this dissertation. A number of machine learning methods are available which can be used to build the SDP models. Only a systematic review can provide an overview of the commonalities and the differences between these studies. Our work performs a review study of the practice of the soft computing meth