中文摘要：

针对传统软件缺陷预测方法在预测面向对象的软件缺陷时存在的不足,提出一种基于流形学习的面向对象的软件缺陷预测模型。结合拉普拉斯特征映射法和分类方法,利用拉普拉斯特征映射法,对待预测的软件属性度量数据进行降维处理,得到其低维特征;利用分类算法,从低维特征数据中预测软件存在的缺陷。实验结果表明,该方法有效提高了面向对象的软件缺陷预测精度,也提高了算法的执行效率。

英文摘要：

To overcome the faults of the traditional software defect prediction methods in predicting object oriented software de- fects, a software defect prediction model for the object oriented software based on the manifold learning was proposed through combining the Laplacian eigenmaps and classification algorithms. The dimension of software metrics attributes data was reduced using the Laplacian eigenmaps to obtain the low dimensional features. The defects existing in the software were predicted from low dimensional data using classification algorithms. Experimental results show that the prediction accuracy is effectively im- proved usin~ the proposed algorithm, and the execntin~ effieionrv r~f thp nl~nrlthm~ ie al ,1 A