An Empirical Evaluation of the Tobit Model on Software Defect Prediction
2016 4th Intl Conf on Applied Computing and Information Technology/3rd Intl Conf on Computational Science/Intelligence and Applied Informatics/1st Intl Conf on Big Data, Cloud Computing, Data Science & Engineering (ACIT-CSII-BCD)(2016)
摘要
In project management, project plan is made based on the prediction results of the project. Predicting the number of defects is one of important prediction. To enhance the prediction accuracy of the number of defects, many studies proposed various prediction models. The model is built using a dataset collected in past projects, and the number of defects is predicted using the model and the data of the current project. Datasets sometimes have many data points where the dependent variable, i.e., the number of defects is zero. When a multiple linear regression model is made using the dataset, the model may not be built properly. To build proper model, we use the Tobit model as software defect prediction. The model assumes that the range of a dependent variable is limited, e.g., the minimum value of the variable is zero, and the model is built based on the assumption. In the experiment, we applied the regression model based on ordinary least squares and the Tobit model to fault prediction. Also, we evaluated models applied log-transformation. In the experiment, the Tobit model applied log-transformation was the highest accuracy in the models. Median BRE of the model was 14% improvement, and Pred25 was 7% improvement, compared with other models.
更多查看译文
关键词
Fault prediction,censored data,log-transformed,linear regression
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络