基于GA-tBP模型的城市道路交通量预测

针对城市道路交通量时间差异性强的问题，提出了一种使用遗传算法(genetic algorithm,GA)优化BP(back propagation)神经网络并考虑时间特性的模型(GA-tBP)预测道路交通量。从月、周、日3个维度统计分析了历史交通量数据存在的差异，得出影响城市道路交通量的时间因素。以江西省南昌市赣江中大道交通量为例，对未来24 h内每5 min交通量进行预测，对比分析了不同模型在预测精度和学习能力等方面的表现。研究结果表明，同不考虑时间特性的模型及其他传统预测模型相比较，GA-tBP模型的预测效果最好，其均方误差(mean squared error, MSE)及平均绝对误差(mean absolute error, MAE)分别为14.87 veh/5 min和2.44 veh/5 min,故该模型具有可行性和有效性。

Urban road traffic volume prediction based on GA-tBP model

Aiming at the problem of strong time difference of urban road traffic volume, a GA-BP neural network model considering time characteristics is proposed to predict road traffic volume. The differences in historical traffic volume data are analyzed from the three dimensions of month, week and day, so as to obtain the time factors affecting urban road traffic volume. Taking the traffic volume of Ganjiangzhong Avenue in Nanchang City, Jiangxi Province as an example, the traffic volume is predicted to be every 5 minutes in the next 24 hours. The performance of different models in prediction accuracy and learning ability is compared and analyzed. The results show that compared with the model without considering the time characteristics and other traditional prediction models, GA-tBP model has the best prediction effect, and its MSE and MAE values are 14.87 veh·5min-1 and 2.44 veh·5min-1, respectively. Therefore, the model is feasible and effective.