01593362

Component

This paper studies the clustering method to classify the congestion patterns, by utilizing the Traffic Performance Index (TPI) data and the TPI curves in Beijing. By analyzing the traditional clustering method in an existing research, two novel clustering methods are presented in this paper. Then, three methods are applied to the TPI data collected from April to June 2014, and the results are compared and evaluated by the measures of the Coefficient of Variations (CV) and the Coverage Ratio (CR). The proposed New Method II is found to be the optimal clustering method, which uses the k-means method, the TPI as the clustering index, the Silhouette measure to determine the optimal category number, and the CV as the weight of the clustering index. Further, the proposed New Method II is validated by conducting the correlation analysis of the clustering results of 2014 and 2015. It is found that TPIs for categories of Monday, Ordinary Weekdays, and Saturday are highly correlated between 2014 and 2015 with a correlation coefficient higher than 0.95. Considering the regular characteristics of the daily traffic conditions, and the strong consistency among different years, the proposed method is helpful in predicting the future traffic conditions.

This paper was sponsored by TRB committee ABJ35 Standing Committee on Highway Traffic Monitoring. Alternate title: Pattern Clustering and Characteristics of Traffic Performance Index in Megacities: Case Study in Beijing.

01584066

16-3320

English

17p

2016

Transportation Research Board 95th Annual Meeting

Digital/other

Figures; References; Tables

Cluster analysis; Correlation analysis; Methodology; Traffic congestion; Traffic forecasting

Beijing (China)

Data and Information Technology; Highways; Operations and Traffic Management

Transportation Research Board Annual Meeting 2016 Paper #16-3320

TRIS, TRB, ATRI

Jan 12 2016 5:28PM