Enhanced Security Metrics Driven Framework for Software Refactoring Prediction and Recommendation Using Adaptive and Attentive Dilation Hybrid Network AADHN

Abstract

Software refactoring is a process that modifies various software artefacts for newlineenhancing the internal structure of the software without impacting the external newlinebehaviour of the software. Commonly, refactoring is employed for improving the newlinequality of the software after a certain number of features have been added. Experts in newlinethe field have analysed with distinct refactoring angles and implemented new models. newlineNevertheless, the network encounters diverse complexities such as the refactoring newlinemodel s diversity, the software system s complexity, and the requirement for smart and newlineautomated models for supporting the developers in making refactoring decisions. For newlinethis purpose, this thesis introduces two significant mechanisms using deep learning. In newlinethe first mechanism, by leveraging a hybrid deep network, a new software refactoring newlineprocess is implemented. In the beginning, the requisite data is garnered from the newlineexisting datasets. Further, this garnered data is provided to the module named feature newlineextraction. In this stage, the ownership measures for entire non-refactored and newlinerefactored data, process, and source code are drawn out. Further, these obtained features newlineare given as input to the AADHN model. This model comprises two techniques such as newlineBiLSTM and DTCN. By leveraging these two techniques, the AADHN model performs newlinethe prediction process. In addition, the developed AADHN employs a new algorithm newlinecalled CIU-GTBO. This algorithm fine-tunes the AADHN model s parameters thus newlineimproving the functionality of the prediction phase newline