基于超欧式距离近邻传播的软件缺陷预测方法

为了进一步提高无标识软件缺陷数据预测的精度，提出了一种基于超欧式距离近邻传播的软件缺陷预测方法。在近邻传播算法中引入密度思想，定义了密度因子和超欧式距离测度概念，设计了密度敏感相似度度量元（即密集度量元），解决了传统近邻传播算法采用欧式距离表示数据相似度，难以有效处理复杂结构数据的不足。该方法应用于无标识软件缺陷数据的预测，并通过三组航空航天软件数据，仿真验证了该方法的有效性，提高了无标识软件缺陷数据预测的精度，为无标识软件缺陷预测提供了一种新的思路。

Software defect prediction based on affinity propagationwith hyper Euclidean distance

In order to improve the accuracy of prediction for unlabeled software defect data, a novel method software defect prediction based on affinity propagation with hyper Euclidean distance, was proposed. Firstly, it being the fact that traditional affinity propagation algorithm using Euclidean distance to represent data similarity, it was difficult to meet the characteristics of global data consistency and cannot effectively deal with the complex data structure. In order to overcome the shortages, the idea of density was introduced, and density factors and hyper Euclidean distance were defined. Meantime, the density sensitive similarity metrics was designed. The new method was used to deal with the unlabeled and complicated software defect data. three data sets are used to verify the effectiveness of the proposed method, and the experimental results show that the proposed method is effective. It improves the prediction accuracy of unlabeled data and provides a practical way for unlabeled software defect prediction.