基于灰色-径向基函数神经网络的交通事故多元预测模型

Multifactor prediction model for traffic accident based on grey-radial basis function neural network

为实现多影响因素作用下的道路交通事故预测,将灰色系统理论和神经网络理论相结合,发挥灰色理论提高可用信息利用率、弱化数据序列波动性的优点及神经网络特有的非线性适应性信息处理能力,提出道路交通事故灰色-径向基函数神经网络多元预测模型,并以某算例进行了不同预测方法结果对比。分析结果表明:与灰色系统预测和径向基函数神经网络预测相比,多元预测模型平均绝对误差、平均绝对百分比误差分别降低50.0%和12.5%,不等系数降低54.5%和16.6%,有效度提高2.7%和0.3%,说明该组合预测能够有效提高系统建模效率与模型精度。

In order to realize the traffic accident prediction under various road factors, a muhifactor prediction model based on grey-radial basis funetion（RBF） neural network was put forward, which combined grey system theory with neural network theory. Grey theory could improve the utilization of available information and weaken the undulation of data sequence. Neural network had the processing ability of nonlinear adaptability information. The muhifactor prediction model was compared with different prediction methods by a case. Analysis result shows that in comparison with grey prediction and RBF neural network prediction, the mean absolute error and mean absolute percent error of muhifaetor prediction model decrease by 50% and 12.5%, the Theil coefficient decreases by 54.5% and 16.6%, the validity increases by 2.7% and 0.3% respectively, so the combined prediction method can improve the efficiency of system modeling and the precision of prediction model.