摘要
为提高飞行事故的预测精度,提出一种基于D-S证据理论的组合预测模型。该模型分别采用时间序列、BP神经网络和最小二乘支持向量机对飞行事故率进行预测,通过对待测年份之前的飞行事故的预测误差分析,计算出相应的基本信任分配函数,并借助D-S证据理论对三种预测模型进行融合,将融合结果作为飞行事故率预测模型的权重,从而得出待测年份的飞行事故预测结果。以美国空军A类飞行事故数据对该组合模型进行验证,结果表明组合预测模型能够较准确地预测飞行事故率,且模型精度优于任何单一预测模型。
This paper puts forward a combination prediction model based on the D-S evidence theory in order to improve the prediction accuracy of flight accidents. In this model,we use time series,BP neural network and least squares support vector machine to respectively predict flight accident rates, and calculate the cor- responding basic trust distribution functions based on the analysis of prediction errors taking place before the year to be predicted. The paper fuses these three models with the help of the D-S evidence theory and takes the fusion model as the weights of the prediction models, and calculates the flight accident errors in the year to be predicted. The simulative result shows that the proposed combined prediction model is more accurate on flight accident prediction than any single of the three models.
出处
《安全与环境工程》
CAS
2015年第3期117-121,共5页
Safety and Environmental Engineering
基金
国家自然科学基金项目(71401174)
关键词
D-S证据理论
飞行事故率
时间序列
BP神经网络
最小二乘支持向量机
组合预测模型
D-S evidence theory
flight accident rates
time series
BP neural network
least squares ,supportvector machine （LS-SVM）
combination prediction model
