Prediction Of Accuracy On Open Source Java Projects Using Class Level Refactoring
Archana Patnaik, Rasmita Panigrahi, Neelamadhab Padhy · 2020 International Conference on Computer Science, Engineering and Applications (ICCSEA) · 2020
The refactoring approach is used to restructure the software system without altering its external functionalities. Our research objective is to analyze the accuracy of the software metrics using machine learning classifiers. After predicting the accuracy we can refactor the inaccurate source metrics by using different refactoring tools and machine learning algorithms. We have considered 30 code metrics from 3 open source java projects by using different framework like antlr4,junit and oryx at the class level and predict its accuracy using different machine learning classifiers. We have used Gaussian, Bernoulli and Multinomial classifiers to predict the accuracy of the software metrics. Statistical significant test reveals the mean accuracy for the above sources are 33.33%,39%,48.33% respectively. As per the prediction analysis we can correct the fault and improve the performance of our source metrics by conducting unit level testing which leads to software refactoring.