Using Source Code Metrics and Ensemble Methods for Fault Proneness\n Prediction

Software fault prediction model are employed to optimize testing resource\nallocation by identifying fault-prone classes before testing phases. Several\nresearchers' have validated the use of different classification techniques to\ndevelop predictive models for fault prediction. The performance of the\nstatistical models are proven to be influenced by the training and testing\ndataset. Ensemble method learning algorithms have been widely used because it\ncombines the capabilities of its constituent models towards a dataset to come\nup with a potentially higher performance as compared to individual models\n(improves generalizability). In the study presented in this paper, three\ndifferent ensemble methods have been applied to develop a model for predicting\nfault proneness. The efficacy and usefulness of a fault prediction model also\ndepends on the source code metrics which are considered as the input for the\nmodel.\n In this paper, we propose a framework to validate the source code metrics and\nselect the right set of metrics with the objective to improve the performance\nof the fault prediction model. The fault prediction models are then validated\nusing a cost evaluation framework. We conduct a series of experiments on 45\nopen source project dataset. Key conclusions from our experiments are: (1)\nMajority Voting Ensemble (MVE) methods outperformed other methods; (2) selected\nset of source code metrics using the suggested source code metrics using\nvalidation framework as the input achieves better results compared to all other\nmetrics; (3) fault prediction method is effective for software projects with a\npercentage of faulty classes lower than the threshold value (low - 54.82%,\nmedium - 41.04%, high - 28.10%)\n