基于Box-Cox转换的集成跨项目软件缺陷预测方法

软件缺陷预测通过预先识别出被测项目内的潜在缺陷程序模块，可以优化测试资源的分配并提高软件产品的质量。论文对跨项目缺陷预测问题展开了深入研究，在源项目实例选择时，考虑了三种不同的实例相似度计算方法，并发现这些方法的缺陷预测结果存在多样性，因此提出了一种基于Box-Cox转换的集成跨项目软件缺陷预测方法BCEL，具体来说，首先基于不同的实例相似度计算方法，从候选集中选出不同的训练集，随后针对这些数据集，进行针对性的Box-Cox转化，并借助特定分类方法构造出不同的基分类器，最后将这三个基分类器进行有效集成。基于实际项目的数据集，验证了BCEL方法的有效性，并深入分析了BCEL方法内的影响因素对缺陷预测性能的影响。

A Box-Cox transformation based ensemble learning approach for cross-project software defect prediction

Software defect prediction can optimize the test resource allocation and improve the quality of software products by predicting the potential defect program modules in advance. We explore the issue of cross-project defect prediction in depth. In instance selection from source projects, we consider three different instance similarity calculation methods and find that the prediction results are diversity. Therefore we propose a Box-Cox transformation based ensemble learning approach named BCEL. In particular, it selects different training sets based on different instance similarity calculation methods. Then it performs Box-Cox transformation based on these datasets and uses a specific classifier to train these base classifiers. Finally it uses an ensemble learning method to combine these base classifiers. Based on datasets from real software projects, we verify the effectiveness of our proposed BCEL approach and analyze the effect of different influencing factors in BCEL