| Autor: |
Zhang, Wei, Ma, Zhen Yu, Lu, Qing Ling, Nie, Xiao Bing, Liu, Juan |
| Zdroj: |
Applied Mechanics and Materials; November 2014, Vol. 687 Issue: 1 p2182-2185, 4p |
| Abstrakt: |
This paper analyzed 44 metrics of application level, file level, class level and function level, and do correlation analysis with the number of software defects and defect density, the results show that software metrics have little correlation with the number of software defect, but are correlative with defect density. Through correlation analysis, we selected five metrics that have larger correlation with defect density. On the basis of feature selection, we predicted defect density with 16 machine learning models for 33 actual software projects. The results show that the Spearman rank correlation coefficient (SRCC) between the predicting defect density and the actual defect density based on SVR model is 0.6727, higher than other 15 machine learning models, the model that has the second absolute value of SRCC is IBk model, the SRCC only is-0.3557, the results show that the method based on SVR has the highest prediction accuracy. |
| Databáze: |
Supplemental Index |
| Externí odkaz: |
|
