Vs Code Extension for Fault Prediction

Abstract

Addressing faults is a complex aspect of the software development process, particularly, when the Program Implementation is considerably lengthy. Accordingly, the emergence of faults upon execution of software entails extra efforts, time and resources to localize the faults, and then initiation of rework to address them. These limitations result in impacting efficiency and making the development process of software cost-intensive. To mitigate the problems faced due to the emergence of faults at the development stage of software, extensive research has been performed in the domain of Software Fault Prediction (SFP). Wherein, SFP pinpoints the emergence of faults in the early development stages of software, thus resulting in an efficient and robust software development process. Software Fault Prediction (SFP) is achieved by classifying modules faulty and non-faulty. Fault data sets are used by models that are applied to the code classes or methods, which are then classified as faulty and non-faulty based on the model predictions. SFP has revolutionized how quality Software is developed with limited time and financial resource constraints. Visual Studio is a powerful editor used by developers to accomplish the entire software development cycle in one place. However, despite its extensive use in development, it has not yet been integrated with any tool that specifically operationalizes SFP. Hence, it was considered pivotal to develop a tool for Visual Studio that provides efficient SFP. Our work elaborates the development of a Visual Studio Fault Prediction Add-On that comprises the extraction of code metrics, and implementation of Deep Learning model using BILSTM and Machine learning model using a Support Vector Machine (SVM) to classify modules as faulty and non-faulty. The Visual Studio with SFP technique will remarkably improve the software development process wherein areas of code encompassing errors will be identified well in advance and correspondingly addressing relevant anomaly(s) is focused without any delay. Thus resulting in an economy of effort in terms of time and resources.