Spatial classification of the impact of COVID-19 on bus ridership using smart card data

Abstract

We evaluated the impact of COVID-19 on bus ridership at the grid cell level using smart card data by categorizing ridership time-series in the following three steps; i) computation of dissimilarity indices between two grid cells based on dynamic time warping (DTW) and the difference in the daily average ridership, ii) grid cell mapping based on the dissimilarity indices using multidimensional scaling (MDS), and iii) grid cell classification using k-means clustering. One of the unique features of this study is the combined analysis of time-series trends in ridership, card type usage rates, and macro facility information by zone, which revealed that the decline in the ridership in the zones with visitor attraction facilities can be due to the not a large daily bus user, and that simultaneous school closures can lead to characteristic ridership patterns.