为改善部队汽车分队的安全状况，建立了基于主成分分析与BP神经网络的部队汽车分队安全评价模型。在利用层次分析法建立分队系统的安全评价指标体系的基础上，将专家作为样本，进行物元分析，两者联合确定指标权重，进而得到相对客观的评价样本。对样本提取主成分，使输入变量降维且相互独立，以提高网络训练和预测效果。结果表明，其预测精度优于不采用主成分分析的BP网络模型，且相对误差在4%以内，模型具有可行性。因此，结合了物元分析与主成分分析的BP网络耦合模型能更加客观、准确地评价和预测被评价对象的实际安全状况。%In order to improve the safety situation of military vehicle units,a safety assessment model based on PCA and BP neural network was proposed. On the basis of the scientific safety assessment index system designed through AHP,the matter element analysis was conducted by regarding the experts as samples. By the way of combining AHP with matter element analysis,the weights of each index were determined. Then the raw samples,which were relatively objective,were obtained by calculating the corresponding weight and score for each index. Through ex-tracting the main ingredients from the raw samples,the input variables were reduced and unrelated,by which the BP neural network can be trained and predict much better. The results showed that its calculation accuracy was higher than that of the BP neural network without using the PCA,and the relative errors between actual output and expected output were all less than 4%,so it can be applied to assessing the safety situation of military vehicle units. The coupling model of BP neural network combined with principal component analysis and PCA can be more objec-tive and accurate in the aspect of assessing the actual safety situation of evaluation objects.
展开▼
