BP神经网络在效能评估中的样本训练

Research on Sample Training of BP Neural Network in Efficiency Evaluation

根据WSEIAC模型建立了地面防空导弹武器系统效能评估的指标体系,并依此模型建立与之对应的三层BP神经网络。简要分析了BP算法的实现过程,利用专家打分法和模糊层次法相结合的方式取得该神经网络应用于地面防空导弹武器系统效能评估时的训练样本,并对此神经网络进行学习训练,直至达到精度要求。经验证,该网络在评价地面防空导弹武器系统效能时减少了评估中的人因影响,使评估结果更为科学。

According to the WSEIAC model, an index hierarchy of ground antiaircraft missile armament system＇s efficiency has been developed, and its corresponding three BP nerve network was established. It is briefly concerned with the analysis of the BP algorithm, then through Delphi technique and the fuzzy analytical hierarchy process, several groups of training samples are chosen to train the BP neural networks until the precision meets requirements. It is shown that this BP neural network limits the artificial factors when it was been used to evaluate the ground antiaircraft missile armament system＇s efficiency. And it was concluded that this method is scientific and creditable.