Mutation Score Prediction Based on Traditional Operator of MUJAVA Using Logistic Regression

Sasa Ani Arnomo,Noraini Binti Ibrahim,Andi Maslan

Mutation score is a calculation to measure the quality of the source code by entering mutants. This is done to detect errors so that they can create program perfection. The value obtained comes from the division between the number of dead mutants and the number of non-equivalent mutants. If the mutation scores have a small value, it requires a lot of improvement in writing the program code. In this study, Mutation score will be predicted based on traditional operators of MUJAVA using Logistic Regression. The project source code that has been tested shows selected mutant operator of 1478 operators. A large influence is found on the COI operator with 11,644 weight values and AOIU operators with 10,609 weight values. This means that the COI and AOIU operators influence the increase in mutation scores. The evaluation results using confusion matrix proved that the results of testing logistic regression algorithms have a higher accuracy value than the KNN algorithm, Naïve Bayes, Random Forest, and SVM. The AUC value in the logistic regression algorithm is obtained a perfect value. So, that it can be concluded that the performance of the algorithm for the Iris dataset case can predict the entire test data perfectly.

Volume 11 | 08-Special Issue

Pages: 1230-1238