Developing a Bayesian Network Model Based on a State and Transition Model for Software Defect Detection

Abstract

This paper describes a Bayesian Network model-to diagnose the causes-effect of software defect detection in the process of software testing. The aim is to use the BN model to identify defective software modules for efficient software test in order to improve the quality of a software system. It can also be used as a decision tool to assist software developers to determine defect priority levels for each phase of a software development project. The BN tool can provide a cause-effect relationship between the software defects found in each phase and other factors affecting software defect detection in software testing. First, we build a State and Transition Model that is used to provide a simple framework for integrating knowledge about software defect detection and various factors. Second, we convert the State and Transition Model into a Bayesian Network model. Third, the probabilities for the BN model are determined through the knowledge of software experts and previous software development projects or phases. Last, we observe the interactions among the variables and allow for prediction of effects of external manipulation. We believe that both STM and BN models can be used as very practical tools for predicting software defects and reliability in varying software development lifecycles.