Project 1

To produce the yearly sum bar chart and table, we aggregate the table by the years in the table with the SUM function. The resulting data frame from the aggregate function will have a column with the numerical year and a column containing the sum for all the days in that year. For organization, I renamed these columns to “Year” and “Riders” respectively. To generate the bar plot, I used the geom_bar function within the ggplot2² package by using the “Year” column as the x and “Riders” as the y axis. Since the yearly sums are a large quantity, I scaled the “Riders” column by 1,000,000 for this visualization only to make it easier to read and reflected that on the y-axis label. The resulting data frame is also used to produce the table underneath the chart but with a few changes for appearance. To make the table easier to read, I used the prettyNum function to add commas within the “Riders” column and casted the “Year” column as an integer so that the column is rendered correctly.

To produce the month bar chart and table, a similar workflow to the one used to produce the yearly sum is used. The key difference is by making a subset of the desired year before performing the same aggregation on the twelve months found within a year using the month() function found in lubridate³. The same ggplot² function is used as in the year bar chart but with the “Riders” column scaled by 1,000. The table is exactly like the table found in the yearly sum minus the integer casting of the date column. For the day of week visualizations, the same steps are used as the month visualizations with the only difference being the aggregation function using the days of the week provided by the wday() function from lubridate³.

Unlike the month and day of week visualizations, the daily visualization is much simpler. The workflow is like the month visualization minus the aggregation and scaling of the “Riders” column as the data frame already represents the day-to-day totals. To produce a good-looking table, the date within the data frame is formatted using the format() function to a year-month-day format. Since there are numerous days to show within the table, the data shown within the table is a subset between two date ranges selected by a dateRangeInput() widget in the Shiny application.

The UI and server functionality is handled by the Shiny package. To layout the three different Blue Line stops, a fluidPage is used with three columns of size 4. Included within each column, one can find a header representing the stop name, a selectInput() widget with the options being each year between 2001-2021, a radioButtons() widget with the different scales as options, a plotOutput() widget for the plot, and a tableOutput() widget for the table. In addition, there is a conditional statement to show a dateRangeInput() widget above the table if the radioButtons() widget is selected to “Day” and another conditionial that hides all other options other than “Year” if the selected year range is “2001-2021”. The server runs the R code mentioned above using the corresponding widgets as input.

Interesting Finds

The most outstanding trend found within all three Blue Line stops is the decline in ridership counts due to the COVID-19 pandemic in 2020. This trend can be seen in the initial graphs showing the yearly totals from 2001-2021 with the 2020 and 2021 bars are significantly decreased compared to the previous years. Selecting the month view for the year 2020 for each stop, we can see that the lockdown affected the Blue Line ridership in March of 2020. Using the day view for the year 2020 and narrowing the selection to March 1^st to March 31^st, we can see that the decrease was sudden but not instantaneous. The decline starts on March 14^th and settles around March 20^th in all three graphs. One thing to notice in this decline is that the rate of the decline is different between all three graphs. Most notably, the UIC-Halsted stop had a steep decline while the O’Hare Airport stop had a smoother decline. This could be since UIC students were no longer commuting to campus on the specific day that it closed while O’Hare Airport flyers still flew on essential flights while the lockdown was being put into place within other states and countries. There are no other declines in ridership on the CTA Blue Line as severe as the COVID-19 pandemic caused that can be found within the years of this data set.

Looking back at years before the COVID-19 pandemic, we can find individual trends within each of the stops displayed. Using the day-of-week view at UIC-Halsted, we can see that this Blue Line stop is most used during weekdays and not the weekends. The weekends notice a significant drop in ridership, about a quarter of a weekday ridership, in each year prior to the pandemic. Using the same view for the Forest Park stop, the same trend can be found as in the UIC-Halsted stop. The weekends show a noticeable drop in ridership, while not as large as in UIC-Halsted, at the Forest Park stop. Interestingly, this trend is not found at the O’Hare stop as ridership is mostly consistent throughout the week. We can hypothesize that the UIC-Halsted and Forest Park stops are mostly used by students and typical workers that go to classes and/or work during the week while the O’Hare Airport stop is mostly used by travelers.