基于粒计算的交通流参数预测

Traffic Flow Parameter Prediction Based on Granular Computing

交通流参数预测是交通流诱导和交通信息发布的重要依据.以信息颗粒为基础数据分析单元,针对以往模糊时间序列模型存在的缺陷,提出一种新方法构建模糊时间序列模型,该方法在挖掘数据内在信息关联的基础上,考虑时间变量影响下分析动态可变的区域间隔长度.此方法主要特点是基于Gath-Geva模糊聚类的时间序列分割,利用模糊分割构造信息颗粒,以信息颗粒为数据单元,通过粒计算分析交通流参数动态变化趋势.实验结果表明,基于粒计算的交通流参数预测可以预测合理的交通流参数置信区间,比传统的参数数值预测可靠度更高,对于交通状态的动态分析具有指导意义.

Traffic flow parameter prediction is the foundation of traffic flow guidance system and traffic information display system. This paper takes information granule as basic analysis unit, and provides a brand new approach to build fuzzy time series model aiming to neutralize original model＇s defects. Based on digging internal correlations of the data, this new approach analyzes dynamic regional interval length under the influence of time variables. This approach is characterized as Gath- Geva fuzzy clustering algorithm based on the segmentation of time series, where the information granule serves as data units is analyzed by granular computing, from which the traffic flow parameters＇ fluctuation trends are captured. Experimental results show that traffic flow parameter prediction based on granular computing gives reasonable confidence interval and is more reliable than the original. The new approach is helpful for further traffic state dynamic analysis.