摘要
针对快速路路段日交通流曲线的相似特性,提出根据天气好、坏,工作日、非工作日作为特性向量,设计一种时间序列聚类算法对不同时间尺度下快速路交通流进行预测。结果显示:此聚类算法与预测模型充分利用了历史数据,30 min时间尺度下预测精度较高,平均绝对相对误差最低为3.34%。
Based on the observation of daily similarity of expressway traffic flow data, a new time series clustering algorithm to forecast traffic parameters under different time scales was designed by taking weather and weekday or weekend as classification vectors. The experiment result shows that the time series clustering algorithm and its forecast model make good use of historic data. Its accuracy is quite high under 30-minute time scale with a mean-absolute-relative-error of 3.34 percent.
出处
《交通与计算机》
2008年第5期49-52,共4页
Computer and Communications
关键词
交通流特性
时间序列
时间尺度
聚类
预测
traffic flow characteristics
time series
time scale
clustering
forecast
