Feature Selection Based Software Defect Prediction using Classification Techniques |
Author(s): |
| Suneetha Merugula , GMRIT |
Keywords: |
| Defect Prediction, Classification Techniques, Principle Component Analysis |
Abstract |
|
Software defect prediction aims to determine whether a software module is defect-prone or non-defect prone by building prediction models. The performance of such models is liable to the high dimensionality of the datasets that may include irrelevant and redundant features. Adding feature selection reduces datasets with fewer features and maintains or improves the prediction capability over the original datasets. The performance of these feature selection method is evaluated using three popular classification techniques: Multilayer Perceptron, PART, J48 over four software defect-datasets obtained from the NASA data repository. The performances were measured using Accuracy, F-measure. Our results demonstrated that J48 combined with principle component analysis feature selection gives better accuracy and F-Measure for all datasets. |
Other Details |
|
Paper ID: IJSRDV5I80390 Published in: Volume : 5, Issue : 8 Publication Date: 01/11/2017 Page(s): 391-394 |
Article Preview |
|
|
|
|
