Analyzing the significance of process metrics for TT&C software defect prediction

In the existing studies on software prediction, the most proposed methods are usually assessed over the public datasets like NASA metrics data repository, which include a combination of code metrics merely. Obviously, the process metric is also one of the key factors that affect the defect-proneness of software modules. In this paper, life-cycle based management process metrics set and history change process metrics set have been proposed based on the characteristics of development process. In order to analyze the importance of these different metrics for predicting defects in aerospace tracking telemetry and control (TT&C) software, an improved PSO optimized support vector machine algorithm (PSO-SVM) has been presented and took into application. The experiment results over the actual TT&C projects suggest that the prediction performance can be significance improved if the 2 kinds of process metrics are included in the model.