Classifying defective software projects based on machine learning and complexity metrics

Software defects can lead to software failures or errors at any time. Therefore, software developers and engineers spend a lot of time and effort in order to find possible defects. This paper proposes an automatic approach to predict software defects based on machine learning algorithms. A set of complexity measures values are used to train the classifier. Three public datasets were used to evaluate the ability of mining complexity measures for different software projects to predict possible defects. Experimental results showed that it is possible to min software complexity to build a defect prediction model with a high accuracy rate.