基于后验概率SVM的仿真模型验证方法  

Validation of simulation model based on posterior probability for support vector machine

作　　者：吴晓燕[1] 吴静[1] 滕江川[1] 陈永兴[1] 

机构地区：[1]空军工程大学导弹学院,陕西三原713800

出　　处：《计算机应用研究》2011年第7期2550-2552,2564,共4页Application Research of Computers

基　　金：陕西省自然科学基础研究计划资助项目(2007F40);总装备部"十一五"武器装备预先研究资助项目(513040104)

摘　　要：针对复杂系统建模与仿真中模型验证的特点,提出基于后验概率支持向量机(support vector machine,SVM)的仿真模型验证方法。采用基于误差分析的特征提取方法提取各可选仿真模型和实际系统输出的特征向量,利用各可选模型的特征向量及相应的模型标号组成训练集建立后验概率SVM模型,将实际系统的特征向量作用于后验概率SVM模型,依据其概率输出判别可选模型相对于实际系统的有效性。以某型飞行器制导控制系统仿真模型验证为例,证明了该方法的可行性和有效性。To validate simulation models of complex system effectively,this paper proposed the validation methodology of posterior probability for support vector machine（SVM）.Extracted the output feature vectors of alternative simulation models and real system by error analysis feature extraction method.By learning the feature vectors of alternative simulation models,established the posterior probability for SVM model.Once the feature vector of the real system was offered to the model,generated the probability output.Judged the validity of each alternative simulation model relative to the real system from the corresponding probability output.The model with the highest probability could be retained as a valid model.Applied the method to the model validation of an aircraft guidance and control system.The result shows the reasonability and validity of the proposed approach.

关 键 词：模型验证 支持向量机 后验概率 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]