Evaluating the Impact of Data Transformation Techniques on the Performance and Interpretability of Software Defect Prediction Models

Figure 3

The overall performance differences on all six performance evaluation indicators across all datasets between all transformed models and original models. E.g., Abscissa L-O represents the difference between the performance indicators of the SDP model constructed by the transformed datasets using the log transformation method minus the performance indicators of the SDP model constructed by the original datasets.