针对传统态势评估方法权值确定困难、大规模数据处理和特征提取能力不足的问题,结合当前空战数据特征,将深度置信网络(DBN)应用于近距空战态势评估。通过密度峰值算法对空战特征数据进行聚类分析,并结合态势函数和专家判读进行修正,建立标准空战态势样本库;以重构误差和测试错误率为基础,建立网络拓扑结构和最优参数确定方法,提高模型的训练效率,并通过样本数据,对模型进行训练和验证。实验表明,模型态势分类正确率达到92.7%,模型运行时间满足应用需求,实例评估结果与客观态势一致性强。 In view of the difficulty of weight determination, large-scale data processing and feature extraction of traditional situation assessment methods, the depth confidence network (DBN) is applied to the situation assessment of short-range air combat combined with the characteristics of current air combat data. Based on the density peak algorithm, the air combat characteristic data is clustered, and combined with the situation function and expert interpretation to correct and establish the standard air combat posture sample database. Based on the reconfiguration error and the test error rate, the network topological structure and the optimal parameters are established Method to improve the training efficiency of the model, and through the sample data, the model training and verification. Experiments show that the correct rate of model situation classification reaches 92.7%, the model running time meets the application requirements, and the result of case evaluation is in good agreement with the objective situation.