摘要
对于可扩展有限状态机(EFSM)规格说明,影响路径测试数据生成成本的因素很多,它们之间可能存在着相互关联,对测试数据生成成本的影响可能是线性或非线性的,因此建立多元线性回归预测模型和BP神经网络非线性预测模型,对EFSM路径测试数据生成进行效率—因素分析。具体而言,将路径长度、路径上变量数等因素作为自变量,测试生成成本看作因变量,建立多元线性回归模型。对于BP神经网络模型,考虑到因素间可能存在关联,首先采用主成分分析(PCA)确定影响测试生成成本的主要因素,然后应用BP神经网络建立测试生成效率主要因素分析模型,对EFSM测试生成成本进行预测。实验结果表明:BP神经网络比多元线性回归更适合作为EFSM路径测试生成效率因素分析模型,对EFSM路径测试生成成本进行非线性预测。
For Extended Finite State Machine （EFSM） specifications, there are many factors that affect the test generation cost, and perhaps, these factors are related to each other. In addition, the test generation cost may be influenced by the factors in a linear or non-linear way. Therefore, a multiple linear regression predictive model and a BP neural network predictive model were established respectively, to conduct the analysis of effficiency-factors for path-oriented test generation on EFSM. According to the multiple linear regression model, the path factors such as the length of path, the number of variables on the path were treated as independent variables, and the test generation cost was considered as a dependent variable when the model was applied to predict the cost. For the BP neural network model, considering that the factors are possibly correlative, Principal Component Analysis （PCA） was firstly utilized to find the principal factors that affect the cost. After that, the BP neural network was applied to construct the analysis model of efficiency-factors. The experimental results demonstrate that compared to the multiple linear regression model, the BP neural network is more suitable for the analysis model of efficiency-factors to predict nonlinearly the test generation cost for paths in EFSM.
出处
《计算机应用》
CSCD
北大核心
2013年第A02期229-234,共6页
journal of Computer Applications
基金
国家自然科学基金资助项目(61073035
61170082)
教育部新世纪优秀人才支持计划项目(NCET-12-0757)
关键词
BP神经网络
多元线性回归
主成分分析
可扩展有限状态机
测试生成效率
BP neural network
multiple linear regression
Principal Component Analysis （PCA）
Extended Finite State Machine （EFSM）
test generation efficiency
