Fig 1.
Ten percent and one percent subset of weekend trip map in 2013.
Fig 2.
Neighboring stations and trips for the given distance.
Fig 3.
Comparison of Divvy daily usage between two years.
Fig 4.
Subscribers’ biking behavior at different time periods.
Fig 5.
Customer’ biking behavior at different time periods.
Fig 6.
Median of trip duration for males and females in 2013 and 2014.
Fig 7.
(a) Flow clusters for morning peak hours on weekdays. (b) Flow clusters for afternoon peak hours on weekdays. (c) Customers’ travel patterns during weekend. (d) Subscribers’ travel patterns during weekend. Background color represents the flow-in density. Red color means trips converge while green color means trips flow away. Group 1 represents 2013 and Group 2 represents 2014.
Fig 8.
Demands for dock and bike that exceed the hypothesized service capacity.
Counts above zero represent the total number of bikes that are not able to check in due to full docks. Counts below zero represent the total number of failed attempts to check out bikes due to empty docks. Colors represent different time windows.
Fig 9.
Spatiotemporal clusters for bike and dock over-demand.
Stations were classified into five classes. Each cluster has its own temporal use pattern. Red curve represents the over check-in pattern and green curve represents the over check-out pattern.
Table 1.
User profile of each cluster when the modeled over-demand happened in 2013 and 2014.
Fig 10.
The over-demand directional profile for stations in Cluster B in the evening.
Fig 11.
The stations had significantly more check-out records than check-in records during both morning and afternoon peaks in Cluster C and D.
Table 2.
The average area of certain land use type in station buffers.
Unit of the area is square meter. Numbers in the parenthesis represent the percentage of land use of certain kind among five clusters.