关键词: 城市交通, 出行规律, 信息熵, 重复性, 周期性, K-Means++

Abstract: Existing literature on the regularity of bus passengers' travel behavior has primarily focused on measuring repeatability and periodicity separately. However, it is recognized that repeatability and periodicity coexist in the travel patterns of individuals. To address this, this paper proposes a method that combines repeatability and periodicity to measure the regularity of bus passengers. Daily activity sequential chains of bus passengers are utilized to identify individual travel patterns. And the information entropy model is improved to measure repeatability and the periodic detection function is improved to measure periodicity. By combining these measures, a K-means++ clustering method is developed for measuring the regularity of bus passengers. To evaluate the method, we obtain data from October to December 2019 in Haining City, Zhejiang Province, China. Four information entropy indices are presented to measure regularity, including temporal repeatability, spatial repeatability, temporal periodicity, and spatial periodicity. A total of 71080 bus passengers were classified into three groups featured with high repeatability and periodicity, low repeatability and high periodicity, low repeatability and periodicity. And we compare the frequency distribution of bus passengers with different departure times across the three groups. The results show that by combining repeatability and periodicity, an additional 21692 passengers, which account for 30.52% of all passengers, can be classified into an interpretable group. We also provide visualizations of travel patterns for the most regular passengers in each group, further illustrating the performance of the proposed measure. This paper demonstrates that combining repeatability and periodicity is a valuable approach for accurately identifying and understanding the individual travel demand of bus passengers.

Key words: urban traffic, travel regularity, information entropy, repeatability, periodicity, K-Means++