Data Visualisation

Last updated on 2024-06-14 | Edit this page

Overview

Questions

  • How can I use Python tools like Pandas and Plotly to visualize library circulation data?

Objectives

  • Generate plots using Python to interpret and present data on library circulation.
  • Apply data manipulation techniques with pandas to prepare and transform library circulation data into a suitable format for visualization.
  • Analyze and interpret time-series data by identifying key trends and outliers in library circulation data.

For this module, we will use the tidy (long) version of our circulation data, where each variable forms a column, each observation forms a row, and each type of observation unit forms a row. If your workshop included the Tidy Data episode, you should be set and have an object called df_long in your Jupyter environment. If not, we’ll read that dataset in now, as it was provided for this lesson.

PYTHON

#import if it is already not
import pandas as pd
df_long = pd.read_pickle('data/df_long.pkl')

Let’s look at the data:

PYTHON

df_long.head()
branch address city zip code ytd year month circulation
date
2011-01-01 Albany Park 5150 N. Kimball Ave. Chicago 60625.0 120059 2011 january 8427
2011-01-01 Altgeld 13281 S. Corliss Ave. Chicago 60827.0 9611 2011 january 1258
2011-01-01 Archer Heights 5055 S. Archer Ave. Chicago 60632.0 101951 2011 january 8104
2011-01-01 Austin 5615 W. Race Ave. Chicago 60644.0 25527 2011 january 1755
2011-01-01 Austin-Irving 6100 W. Irving Park Rd. Chicago 60634.0 165634 2011 january 12593

Plotting with Pandas


Ok! We are now ready to plot our data. Since this data is monthly data, we can plot the circulation data over time.

At first, let’s focus on a specific branch. We can select the rows for the Albany Park branch:

PYTHON

albany = df_long[df_long['branch'] == 'Albany Park']

PYTHON

albany.head()
branch address city zip code ytd year month circulation
date
2011-01-01 Albany Park 5150 N. Kimball Ave. Chicago 60625.0 120059 2011 january 8427
2012-01-01 Albany Park 5150 N. Kimball Ave. Chicago 60625.0 83297 2012 january 10173
2013-01-01 Albany Park 5150 N. Kimball Ave. Chicago 60625.0 572 2013 january 0
2014-01-01 Albany Park 5150 N. Kimball Ave. Chicago 60625.0 50484 2014 january 35
2015-01-01 Albany Park NaN NaN NaN 133366 2015 january 10889

Now we can use the plot() function that is built in to pandas. Let’s try it:

PYTHON

albany.plot()
Line plot of zip code, ytd, year, and circulation numbers over time from the albany DataFrame

That’s interesting, but by default .plot() will use a line plot for all numeric variables of the DataFrame. This isn’t exactly what we want, so let’s tell .plot() what variable to use by selecting the circulation column.

PYTHON

albany['circulation'].plot()
Line plot of the Albany Park branch circulation showing a big drop from 2013 to 2014.

The significant drop in circulation in March 2020 is likely due to the COVID-19 pandemic, which caused widespread temporary closures of public spaces, including libraries.

The drop from 2012 through part of 2014 corresponds to the reconstruction period of the Albany Park Branch. The original building at 5150 N. Kimball Avenue was demolished in 2012, and a new, modern building was constructed at the same site. The new Albany Park Branch opened on September 13, 2014, at 3104 W. Foster Avenue in the North Park neighborhood of Chicago. More details about this renovation can be found on the Chicago Public Library webpage: Chicago Public Library - Albany Park.

Use Pandas for More Detailed Charts


What if we want to alter the axis labels and the title of the graph? Pandas’ built-in plotting functions, which are backended by Matplotlib, allow us to customize various aspects of a plot without needing to import Matplotlib directly.

  • We can pass parameters to Pandas’ .plot() function to add a plot title, specify a figure size, and change the color of the line.
  • Additionally, we can directly set the x and y axis labels within the .plot() function.

PYTHON

albany['circulation'].plot(title='Circulation Count Over Time', 
                                figsize=(10, 5), 
                                color='blue', 
                                xlabel='Date',
                                ylabel='Circulation Count')
Line plot of the Albany Park branch circulation with matplotlib styles applied.

Changing plot types

What if we want to use a different plot type for this graphic? To do so, we can change the kind parameters in our .plot() function.

PYTHON

albany['circulation'].plot(kind='area', 
                            title='Circulation Count Area Plot at Albany Park', alpha=0.5, 
                            xlabel='Date',
                            ylabel='Circulation Count')
Area plot of the Albany Park branch circulation.

We can also look at our circulation data as a histogram.

PYTHON

albany['circulation'].plot(kind='hist', bins=20, 
                            title='Distribution of Circulation Counts at Albany Park',
                            xlabel='Circulation Count')
histogram of the Albany branch circulation.

Use Plotly for interactive plots


Let’s switch back to the full DataFrame in df_long and use another plotting package in Python called Plotly. First let’s install and then use the package.

PYTHON

# uncomment below to install plotly if the import fails. 
# !pip install plotly
import plotly.express as px

Now we can visualize how circulation counts have changed over time for selected branches. This can be especially useful for identifying trends, seasonality, or data anomalies. We willfirst create a subset of our data to look at branches starting with the letter ‘A’. Feel free to select different branches. After subsetting, we will sort our new DataFrame by date and then plot our data by date and circulation count.

PYTHON

# Creating a line plot for a few selected branches to avoid clutter
selected_branches = df_long[df_long['branch'].isin(['Altgeld',
 'Archer Heights',
 'Austin',
 'Austin-Irving',
 'Avalon'])]
selected_branches = selected_branches.sort_values(by='date')

PYTHON

fig = px.line(selected_branches, x=selected_branches.index, y='circulation', color='branch', title='Circulation Over Time for Selected Branches')
fig.show()

Here is a view of the interactive output of the Plotly line chart.

One advantage that Plotly provides over Matplotlib is that it has some interactive features out of the box. Hover your cursor over the lines in the output to find out more granular data about specific branches over time.

Bar plots with Plotly

Let’s use a barplot to compare the distribution of circulation counts among branches. We first need to group our data by branch and sum up the circulation counts. Then we can use the bar plot to show the distribution of total circulation over branches.

PYTHON

# Aggregate circulation by branch
total_circulation_by_branch = df_long.groupby('branch')['circulation'].sum().reset_index()

# Create a bar plot
fig = px.bar(total_circulation_by_branch, x='branch', y='circulation', title='Total Circulation by Branch')
fig.show()

Here is a view of the interactive output of the Plotly bar chart.

Plotting with Pandas

  1. Load the dataset df_long.pkl using Pandas.
  2. Create a new DataFrame that only includes the data for the “Chinatown” branch.
  3. Use the Pandas plotting function to plot the “circulation” column over time.

PYTHON

import pandas as pd
df_long = pd.read_pickle('data/df_long.pkl')
chinatown = df_long[df_long['branch'] == 'Chinatown']
chinatown['circulation'].plot()
image showing the circulation of the Chinatown branch over ten years
Chinatown plot

Modify a plot display

Add a line to the code below to plot the Uptown branch circulation including the following plot elements:

  • A title, “Uptown Circulation”
  • “Year” and “Circulation Count” labels for the x and y axes
  • A green plot line

PYTHON

import pandas as pd
df_long = pd.read_pickle('data/df_long.pkl')
uptown = df_long[df_long['branch'] == 'Uptown']

PYTHON

uptown['circulation'].plot(title='Uptown Circulation', 
                                color='green', 
                                xlabel='Year',
                                ylabel='Circulation Count')
image showing the circulation of the Uptown branch with labels
Uptown plot

Plot the top five branches

Modify the code below to only plot the five Chicago Public Library branches with the highest circulation.

PYTHON

import plotly.express as px
import pandas as pd
df_long = pd.read_pickle('data/df_long.pkl')
total_circulation_by_branch = df_long.groupby('branch')['circulation'].sum().reset_index()

top_five = total_circulation_by_branch.___________________

# Create a bar plot
fig = px.bar(top_five._______, x='branch', y='circulation', width=600, height=600, title='Total Circulation by Branch')
fig.show()

PYTHON

total_circulation_by_branch.sort_values(by='circulation', ascending=False)
df_long = pd.read_pickle('data/df_long.pkl')
total_circulation_by_branch = df_long.groupby('branch')['circulation'].sum().reset_index()

top_five = total_circulation_by_branch.sort_values(by='circulation', ascending=False)

# Create a bar plot
fig = px.bar(top_five.head(), x='branch', y='circulation', width=600, height=600, title='Total Circulation by Branch')
fig.show()
a bar plot of the top five branch circulation figures
Top five circulation branches

Key Points

  • Explored the use of pandas for basic data manipulation, ensuring correct indexing with DatetimeIndex to enable time-series operations like resampling.
  • Used pandas’ built-in plot() for initial visualizations and faced issues with overplotting, leading to adjustments like data filtering and resampling to simplify plots.
  • Introduced Plotly for advanced interactive visualizations, enhancing user engagement through dynamic plots such as line graphs, area charts, and bar plots with capabilities like dropdown selections.