Exploring Several Strategies to Improve Software Quality Prediction

摘要

This paper investigates an approach for using feature selection and data sampling together to deal with two problems, which often arise in the software quality modeling process, namely high dimensionality of software metric data and class imbalance often found in these data sets. Six commonly used feature ranking techniques and six sampling methods are examined. A case study is performed on three software measurement data sets obtained from the PROMISE repository. The empirical results demonstrate that different feature ranking techniques and sampling methods may significantly affect the performance of classification models.