基于生命周期的软件缺陷预测技术

Lifecycle-based software defect prediction technology

为保证软件可靠性和软件质量,在基于软件开发周期的基础上,提出了一种利用PCA-BP模糊神经网络的软件缺陷预计方法。针对影响软件可靠性的各种因素,依据相关的标准,结合工程实践,选取了影响软件可靠性的度量元。收集了实际工程中的一类飞行控制软件的度量数据,利用提出的模型进行缺陷预测,并将预测结果与传统的BP神经网络模型计算的结果进行了对比。对比结果表明,与基于BP神经网络的预测方法相比较,结合了主成分分析方法的PCA-BP神经网络预测方法具有更快的收敛速度和更高的预测准确度。

To ensure software reliability and software quality, lifecycle-based software defect prediction models using PCA-BP neural network is presented. Metrics are selected according to the relevant standard and the experience of engineering practice. Special flight command software＇s metrics are measured and some prediction results are presented. The prediction results show that PCA-BP neural network has better astringency and precision than BP neural network.