Software Defect Prediction Using Artificial Neural Network

Abstract

Defect severity assessment is the most crucial step in major companies and organizations where the complexity of the software is growing at an exponential rate. One of most active research areas in software engineering is software defect prediction (SDP). SDP is a technique for envisioning software defects. Early defect discovery during the software development life cycle (SDLC) results in early repairs and ultimately on-time delivery of maintainable software, which pleases the client and fosters his confidence in the development team. By consistently predicting bugs, removing bugs, and identifying defect modules, the software industry aims to improve the quality of its products. The requirement for high-quality and affordable software that can be maintained is growing as the need for automated online software systems rises daily. Predicting software defects is one of the major goals of the quality assurance process, which improves quality while reducing costs by reducing overall testing and maintenance activities. Artificial Neural Networks (ANN), one of the commonly utilized machine learning approaches, are used in the majority of suggested frameworks and models for defect prediction. Using Artificial Neural Networks (ANN) approach, Laverberg-Marquardt (LM) and Bayesian Regularization (BR) we can simply predict software defect. To use the MATLAB simulation tool, a framework is built and used to the NASA software dataset being considered for performance study. By considering the performance I can select which ANN method algorithm is more perfect in predicting software defect. This study will benefit the researchers and serve as a benchmark for future improvements, analyses, and evaluations. An experimental investigation demonstrates that the suggested approach can offer superior performance for predicting software defects.