关键词: 城市交通, 信息颗粒, 模糊聚类, 交通流参数, 粒计算

Abstract:

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.

Key words: urban traffic, information granule, fuzzy clustering, traffic flow parameters, granular computing