作战仿真中基于BP-Adaboost目标威胁估计研究  

The Research on Target Threat Assessment Based on BP-Adaboost in Operational Simulation

作　　者：张峰[1] 胥文[1] 吴东岩 陈振朋 ZHANG Feng;XU Wen;WU Dongyan;CHEN Zhenpeng(School of Aviation Operations and Services,Aviation University of Air Force,Changchun 130022,China)

机构地区：[1]空军航空大学航空作战勤务学院

出　　处：《系统仿真技术》2019年第3期180-183,共4页System Simulation Technology

摘　　要：对空中来袭目标进行威胁估计是作战中指挥决策的重要部分,是在目标态势的基础上,通过对目标数据量化处理而进行的综合估计,为指挥员进行兵力部署和火力分配提供重要依据。针对单一神经网络预测估计时网络结构选择困难、泛化能力差的缺点,提出了采用BP神经网络作为弱预测器,通过Adaboost进行集成学习,从而建立BP-Adaboost强预测器目标威胁估计模型。通过对不同态势情况下的样本数据进行学习,更新神经网络权值,生成BP-Adaboost强预测器。结果表明,该方法的预测误差明显优于BP 、PSO-SVM和Elman-Adaboost算法。Target threat assessment is an important part of command and decision-making in combat.It is a comprehensive evaluation based on the target situation through quantitative processing of target data,which can provide significant evidence for the commander to make firepower allocation.To improve the shortcomings of difficult selection of network structure and poor generalization ability in single neural network prediction and evaluation,this paper uses BP neural network as a weak predictor and integrates learning through Adaboost to establish a target threat assessment model of BP-Adaboost strong predictor.By learning the sample data in different situations,the weights of the neural network are updated and the strong predictor of BP-Adaboost is generated.The experiment indicates that the prediction error of this method has higher target threat prediction accuracy than the normal BP、PSO-SVM and Elman-Adaboost algorithm.

关 键 词：BP神经网络 集成学习 威胁估计 辅助决策 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]