Change-Proneness of Object-Oriented Software Using Combination of Feature Selection Techniques and Ensemble Learning Techniques
| Autor: | N. L. Bhanu Murthy, Lov Kumar, Sangeeta Lal, Anjali Goyal |
|---|---|
| Rok vydání: | 2019 |
| Předmět: |
Object-oriented programming
Source code Wilcoxon signed-rank test Computer science business.industry media_common.quotation_subject 05 social sciences Univariate 020207 software engineering Feature selection 02 engineering and technology computer.software_genre Ensemble learning Software metric Software 0502 economics and business 0202 electrical engineering electronic engineering information engineering Data mining business computer 050203 business & management media_common |
| Zdroj: | ISEC |
| Popis: | Change-prone modules are characterized as the programming parts in the source code which have high probability to alter in the future. Change-proneness prediction helps software testers to streamline and concentrate their testing assets on the modules which have a higher probability of alteration. In this work, we perform an empirical study of 11 feature selection techniques to identify the suitable set of source code metrics, out of 21 metrics, for change-proneness prediction. We first proposed a source code validation framework that includes Wilcoxon signed rank test, univariate logistic regression analysis, cross correlation analysis, multivariate linear regression stepwise forward selection. We then compared the results of proposed software metrics validation framework (PFST) with 10 feature selection techniques. The selected features are then used for predicting change-proneness using 18 machine learning algorithms. Experimental results show that the prediction model built using source code metrics obtained from PFST outperforms other techniques. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|