计算机软件质量综合评判的人工神经网络模型  

Comprehensive evaluation model for computer software quality using artificial neural network

作　　者：楼文高[1] 姜丽[2] 孟祥辉[2] 

机构地区：[1]上海理工大学出版印刷与艺术设计学院,上海200093 [2]上海理工大学管理学院,上海200093

出　　处：《上海理工大学学报》2008年第5期479-482,共4页Journal of University of Shanghai For Science and Technology

基　　金：上海市重点学科建设资助项目(P0501)

摘　　要：根据单指标评价标准,通过在规定的区间范围内随机取值生成足够多的训练样本、检验样本和测试样本,遵循误差反传(BP)网络建模基本原则和步骤,建立可靠、有效的软件质量综合评价模型,克服了物元模型、灰色模型等方法存在受人为因素影响及定性评价等缺陷.对6个模拟样本和8个实际软件的评价表明,本文提出的样本生成方法、建模过程能有效避免出现过训练和过拟合现象,建立的BP模型具有较好的泛化能力、可比性和客观性;各评价指标与软件质量等级之间存在明显的非线性关系,可靠性指标对软件质量等影响最大.The sample data for training, verification and testing, in agreement with the singular evaluation index （EI） criterion, were randomly generated. The artificial neural network （ANN） based model with reliability and effectiveness for computer software quality （CSQ） evaluation was established in accordance with the basic principles and steps. The case study shows that the methodology for generating sample data and the process of establishing ANN-model are effective and reliable. The phenomenon such as over-training and over-fitting can be effectively avoided and the established model is of good generalization property. The relationship between the CSQ and EI is nonlinear and the most important index is the reliability.

关 键 词：计算机软件质量 综合评价 人工神经网络 模型 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP311[自动化与计算机技术—控制科学与工程]