【摘要】 为实现降雨条件下高速公路路段行程时间短时预测,掌握恶劣天气下交通信息、提供交通诱导和决策支持,在已获取交通和气象数据基础上应用半距离法估计路段行程时间。并以遗传算法优化的径向基函数(RBF)神经网络和K最近邻非参数回归(KNN)算法为基础,提出1种基于动态权重的行程时间组合预测模型。该组合预测模型的融合权重依据定义的动态误差的变化而持续调整,以保证子模型中精度较高的预测结果对最终结果有较大影响,从而提高预测精度。选取京港澳高速公路湖北省境内军山-武汉南路段,分析该路段降雨条件下行程时间特性,掌握其不同时段和不同降雨强度下行程时间变化规律,并进行预测。结果表明,组合预测模型能有效预测行程时间高峰变化,反应及时且预测精度较高,达到0.98,平均绝对百分误差1.99%;而单一的RBF神经网络和KNN算法的平均绝对百分误差分别为3.40%和2.60%,且拟合程度不如组合预测模型。更多还原

【Abstract】 In order to forecast travel time on freeways under rainfall conditions,collect traffic information in bad weather,and provide guidance for travelers,a combined prediction model based on radial basis function(RBF)neural network and K-nearest neighbor nonparametric regression(KNN)algorithm is developed.The travel time of a road section is estimated by a half-distance method based on actual traffic data and meteorological data.According to dynamic error,weights of the combined forecasting model change to improve forecasting accuracy.Finally,the section of Junshan-Wuhan South Road of Beijing-Hong Kong-Macao freeway in Hubei Province is selected as a case study to analyze characteristics of travel time under the rainfall conditions,and make a forecast.The variation of travel time under different periods and different rainfall intensity is collected.The results show that the precision of the combined forecasting model is 0.98,and the average absolute percentage error is 1.99%,which performs better than RBF neural network and KNN algorithm,of which average absolute percentage errors are 3.40% and 2.60%,respectively.更多还原