基于DBN效能拟合的舰艇编队作战效能敏感性分析

针对传统的舰艇编队作战效能分析方法中存在的对数据利用不充分、对数据完整性要求较高的问题，提出了基于深度学习的效能拟合方法。从最具有代表性的敏感性分析方法Sobol指数法入手，利用深度学习方法优越的特征学习能力，基于深度信念网络（DBN）构建了效能拟合网络，结合无监督预训练和有监督调优实现了网络训练和参数优化，构建出效能拟合模型。将产生的数据应用于效能分析模型并与完全数据条件下的效能分析结果进行对比，验证了所提出的效能拟合模型对于不完全数据下的作战系统敏感性分析的有效性。

Sensitivity analysis of ship formation operational effectiveness based on DBN effectiveness fitting

Aiming at the problem of insufficient data utilization and high requirements for data integrity in the traditional ship formation combat effectiveness analysis analysis method, this paper proposes a performance analysis fitting model based on deep belief network. Start with the most representative sensitivity analysis method-Sobol index method, and then take characteristic learning ability of deep learning, constructing a effectiveness fitting network based on Deep Belief Network(DBN), with network training and parameter optimization combined with unsupervised pre-training and supervised tuning. Finally, the experiments are simulated and analyzed based on the formation of air defense combat. Simulation results verify the applicability and effectiveness of the model.