基于历史频繁模式的交通流预测算法

针对目前交通流预测模型复杂、不支持中长期预测的问题,提出了基于历史频繁模式的交通流预测算法,通过挖掘交通流的历史频繁模式,结合实时交通信息进行交通流预测.使用真实路网获取的浮动车数据进行实验,结果表明该算法支持交通流短时、中长期预测,且中长期预测与短时预测具有同样高的预测精度,受参数影响小.与基于K近邻的非参数回归方法进行比较,结果表明基于历史频繁模式的预测算法的预测性能更稳定,预测误差波动更小.

Traffic flow prediction algorithm based on historical frequent pattern

According to the problem that traffic flow prediction have two defects,have complex models and can’t support the long-term traffic flow prediction,a historical frequent pattern based algorithm is proposed.The algorithm includes two steps.First,mining the frequent patterns of historical traffic flow;second,predict traffic flow combined with real-time traffic information.By using the real probe vehicle data,experimental results show that the algorithm proposed can predict short-term and long-term traffic flow efficient and effective,with high prediction accuracy and robust to parameters.In particular,long-term traffic forecasting with the same high prediction accuracy as short-term traffic forecasting.Finally,compared with the K-NN based nonparametric regression method,higher prediction accuracy and smaller prediction error volatility of this algorithm are shown.