A Systematic Study on Machine Learning Techniques for Predicting Software Faults

Fault tolerance and prediction are some of the interesting quality assurance activities of software engineering. Fault prediction algorithms assist the project manager in identifying the parts of the system that are prone to failure and provide the benefit of reducing the time required to test the entire application. The fault-prediction method attempts to detect faulty software modules so that they are used subsequently in the software development process. There are numerous prediction approaches available in the domain of software engineering for detecting defect-prone blocks. Few of them are stable models, and a few are unstable models, which may not be enough to handle all of the possible scenarios and thus fail to produce the earliest prediction. This paper provides an extensive review of existing predictive models for software defect prediction, identify flaws in existing techniques, and address the need for a new prediction approach in the era of fault prediction. The Classification was commonly used to differentiate among faulty and non-faulty modules. This paper proposes two sub modules for prediction of faults. One submodule to identify the accuracy are implemented and discussed. To achieve better and more accurate results, we intend to build a machine learning model, which can automate the complete process.