The Application of Software Defect Types Prediction Based on Grey Entropy Absolute Relational Analysis

ABSTRACT

In the software development process, reliability prediction can effectively improve its quality. But the small project data sets often restrain the predicting methods of the traditional software detect types, which causes the inaccuracy and unreliability of the predicting results. Grey relational analysis (GRA) is a method which is always used to describe the degree of influence among the factors and suitable for small project data sets. Therefore, on the basis of grey entropy absolute relational analysis (GEARA), we propose a software defect classification prediction method. Firstly, this method chooses an algorithm to carry on analysis and choice for feature attribute, which prediction result has greater relation degree. And then, it detects and removes anomaly project by using anomaly project detection algorithm and gets a set of engineering project. Finally, we use the grey entropy absolute relational analysis to predict software defect type. The simulation experiment indicated that proposed method owe a higher prediction precision than the traditional ones and predicted faster.

Keywords:

Anomaly project, feature subset selection, grey system theory, grey entropy absolute relational analysis, software defect prediction.