Key problems in validation of intelligent models  

作　　者：刘飞 马萍 杨明 王子才 

机构地区：[1]Control and Simulation Center,Harbin Institute of Technology

出　　处：《Journal of Harbin Institute of Technology(New Series)》2009年第3期371-375,共5页哈尔滨工业大学学报（英文版）

基　　金：Sponsored by the National Natural Science Foundation of China(Grant No. 60434010);the Foundation of the Outstanding Youth of Heilongjiang Province ( Grant No. JC200606)

摘　　要：To provide a realistic simulation environment for users,intelligent models have become key components in military simulations. After the analysis of modeling nature of intelligent models,the validation criteria for defining the validation points and validation metrics for measuring the agreements between human experts and intelligent models were presented. Further,such methods as graphical comparison,feature analysis and face validation were discussed according to the characteristics of intelligent models. Based on the validation criteria,validation metrics and validation methods,the intelligent models can be effectively validated,which has been proved in current developed intelligent models.To provide a realistic simulation environment for users, intelligent models have become key components in military simulations. After the analysis of modeling nature of intelligent models, the validation criteria for defining the validation points and validation metrics for measuring the agreements between human experts and intelligent models were presented. Further, such methods as graphical comparison, feature analysis and face validation were discussed according to the characteristics of intelligent models. Based on the validation criteria, validation metrics and validation methods, the intelligent models can be effectively validated, which has been proved in current developed intelligent models.

关 键 词：intelligent models validation criteria validation metrics validation methods 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置] TQ460.7[自动化与计算机技术—控制科学与工程]