authors: Peter Carbonetto, Gao Wang
Here we use the Divvy trip data to examine biking trends over the course of a typical day in Chicago.
We begin by loading a few packages, as well as some additional R functions implemented for this project. The repr
package is used to adjust the dimensions of the figures in the Jupyter notebook.
library(repr)
library(data.table)
library(ggplot2)
source("../code/functions.R")
Following our earlier steps, we use function read.divvy.data
to read the trip and station data from the CSV files.
divvy <- read.divvy.data()
To make it easier to compile statistics by time of day, we convert the "start hour" column to a factor.
divvy$trips <- transform(divvy$trips,start.hour = factor(start.hour,0:23))
Now that start.hour
is a factor, it is easy to create a bar chart showing the total number of departures at each hour. Unsurprisingly, we see very little biking activity at night. Further, the two peaks ("modes") in the bar chart nicely recapitulate the morning and afternoon rush hours.
options(repr.plot.width = 4.75, repr.plot.height = 2.25)
ggplot(divvy$trips,aes(start.hour)) +
geom_bar(fill = "dodgerblue",width = 0.6) +
theme_minimal() +
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank())
However, this bar chart is a bit muddled because it is counting trips during the week and on the weekends. Consider that the bin count $x[h]$ for hour $h$ in the histogram above is a sum of the counts for each day of the week:
$$ \begin{align} x[h] & = \sum_{i\;\in\;\mathsf{DaysOfTheWeek}} x_i[h] \\ & = x_{\mathsf{Mon}}[h] + x_{\mathsf{Tue}}[h] + x_{\mathsf{Wed}}[h] + x_{\mathsf{Thu}}[h] + x_{\mathsf{Fri}}[h] + x_{\mathsf{Sat}}[h] + x_{\mathsf{Sun}}[h] \end{align} $$Note: The math above is embedded in the webpage using MathJax. See here for an excellent reference on MathJax.
Once we plot the counts separately for each the day of the week, the rush-hour trends become more obvious. (Also notice that the rush-hour weeks disappear on Saturday and Sunday.)
options(repr.plot.width = 5, repr.plot.height = 5)
ggplot(divvy$trips,aes(start.hour)) +
geom_bar(fill = "dodgerblue",width = 0.75) +
facet_wrap(~start.day,ncol = 2) +
scale_x_discrete(breaks = seq(0,24,2)) +
theme_minimal() +
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank())
The commuting trends are different at the University of Chicago Divvy station---there isn't much of a morning rush hour. This may be because students and staff don't regularly use the Divvy bikes for commuting.
options(repr.plot.width = 5, repr.plot.height = 5)
ggplot(subset(divvy$trips,from_station_name == "University Ave & 57th St"),
aes(start.hour)) +
geom_bar(fill = "dodgerblue",width = 0.75) +
facet_wrap(~start.day,ncol = 2) +
scale_x_discrete(breaks = seq(0,24,2)) +
theme_minimal() +
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank())
This is the version of Jupyter used to generate these results.
system("jupyter --version",intern = TRUE)
This is the version of R and the packages that were used to generate these results.
sessionInfo()
Exported from analysis/time-of-day-trends.ipynb
committed by Peter Carbonetto on Tue Jul 25 03:57:31 2017 revision 7, a9926b1