摘要
采用K近邻的非参数回归方法对短时交通流量进行了预测,考察了模型中关键因素对预测效果的影响.在4种不同状态向量和预测算法组合下的实验方法比较中,以相邻四个时间间隔的流量和占有率数据作为状态向量,并采用带权重的预测算法取得了良好的效果.将利用K值构造的预测区间用于特殊路况的预测中,得到了明显的改进效果.最后,对非参数回归和神经网络的方法进行了比较,结果表明了非参数回归预测方法的高精度和强移植性.
This paper used the K-NN based nonparametric regression to forecast the short term traffic flow, and analyzed the effect caused by key factors' settings in the model. Within four methods of different setting in traffic state vector and forecast technique, the one which defines traffic flow and occupancy rate in 4 time lags as state vector, and uses weight-added forecast technique has the better simulation results. This paper applied the prediction interval calculated by K to forecast during unconventional road condition, and improved the forecasting results. Finally, nonparametric regression's advantages of high accuracy and strong transplant ability are showed while being compared with neural network.
出处
《系统工程理论与实践》
EI
CSSCI
CSCD
北大核心
2010年第2期376-384,共9页
Systems Engineering-Theory & Practice
基金
国家自然科学基金(70501018,60773124)
上海市自然科学基金(09ZR1420400,09ZR1403000)
上海财经大学“211工程”三期重点学科建设项目
上海市智能信息处理重点实验室开放课题
关键词
短时交通流预测
非参数回归
K近邻
预测区间
状态向量
short-term traffic flow forecasting
nonparametric regression
K nearest neighbor
prediction interval
state vector