Cross-Project Software Refactoring Prediction Using Optimized Deep Learning Neural Network With the Aid of Attribute Selection

Abstract

Cross-project refactoring prediction is prominent research that comprises model training from one project database and testing it for a database under a separate project. While performing the refactoring process on the cross project, software programs want to be restructured by modifying or adding the source code. However, recognizing a piece of code for predicting refactoring purposes is remained to be actual challenge for software designers. To date the entire refactoring procedure is highly dependent on the skills and software inventers. In this manuscript, a deep learning model is utilized to introduce a predictive model for refactoring to highlight classes that need to be refactored. Specifically, the deep learning technique is utilized along with the proposed attribute selection phases to predict refactoring at the class level. The planned optimized deep learning-based method for cross-project refactoring prediction is experimentally conducted on open- source project and accuracy found as 0.9648 as comparison to other mentioned state of the art.