论文部分内容阅读
以英国M42高速公路线圈交通数据为输入,建立预测模型,提出神经网络OD矩阵预测方法.将预测结果与该公路3a号至7号交叉口间各支路上自动车牌识别装置测得的车辆数据进行对比,验证其有效性.解决了以下关键问题:利用线圈数据实现OD矩阵预测的可操作性,该类数据的特殊性是否影响模型构建,通过变异数稳定数据转换能否改善模型性能,能否同时进行单个OD预测.得到基于训练数据平方根代换的最佳计算结果和单个OD预测模型.
Taking the traffic data of the M42 expressway in the UK as input, the prediction model is established and the OD matrix prediction method of neural network is put forward. The vehicle data measured by the automatic license plate recognition device on each branch between the intersection of No. 3a and No. 7 are predicted Contrast and verify the validity of the solution to the following key issues: the use of coil data to achieve the operability of the OD matrix prediction, whether the particularity of this type of data affect the model construction, the stability of the data through the variation of data conversion can improve the model performance, whether At the same time, a single OD prediction is obtained, and the best calculation result based on square root substitution of training data and a single OD prediction model are obtained.