Python for Data Analytics

Module 3: Data Visualization

Data Visualization

Module 3 introduction
Module 3 outline Peer to peer loans

Plotting Visual appeal

Multiple charts
with matplotlib with Seaborn

Column charts Styles Charts on charts

Scatter plots Color palettes Side-by-side

Chained methods Box plots Looping techniques

Histograms

Sean
SeanBarnes
Barnes
 Data Visualization

Plotting with matplotlib

Matplotlib
● A visualization module

import matplotlib

● Customization features:
○ Titles
○ Annotations
○ Colors
○ Axis limits
○ Label formatting

● Hundreds of thousands of lines of

code already written for you

Sean
SeanBarnes
Barnes
 Figure
Figure
● Can customize figure directly
Axes
○ Control the size of the canvas
Legend
○ Set options like background color

Axes (plots)
● Create using different functions:
Grid
○ .plot() Have named arguments:
○ .scatter() ● Data for plotting
○ .hist() ● Data ink for that chart type

Additional chart elements

● .title() ● .annotate() Label
● .xlabel() ● .legend()
● .ylabel()

Sean
SeanBarnes
Barnes
Scenario
🏆 Goal: Develop a state-of-the-art risk management strategy for
providing loans to different communities across the United States

📊 Dataset: Loans from Lending Tree, a peer-to-peer lending platform

🎯 Task: Conducting exploratory data analysis to better understand

the characteristics of loans with different levels of risk

⬜ Develop a report of findings to share with the bank

You
⬜ Develop insightful visualizations to help client understand
Data Analyst different risk profiles

Sean
SeanBarnes
Barnes
Recap: Matplotlib
1. Select & order data using Pandas 3. Use [Link]():

df["grade"].value_counts().sort_index() ● To clean up the output of each code cell

● Multiple times to display multiple plots
2. Stack commands to enhance visualizations

import [Link] as plt

sorted_grades.plot(kind="bar")
[Link]("")
[Link]("Frequency")
[Link]("Frequency of Loan by Grade")

Sean
SeanBarnes
Barnes
 Data Visualization

Colors, grids, & saving plots

Recap: Colors, grids, & saving plots
● To specify color for chart: ● To save an image
"8af133"

sorted_grades.plot(kind="bar", color="purple") [Link]("loan_column_chart.png")

File name
● To use list to give each bar its own color: Any common image format will work:

colors = ["Green", "Lime", … , "Red"] ● .jpeg

● .svg
sorted_grades.plot(kind="bar", color=colors)
● .pdf

● To add grid lines:

[Link](axis="y", color="black", alpha=0.7, linestyle="--")

Specify x or y Color Line style

axes, or both

Sean
SeanBarnes
Barnes
Data Visualization

Text & annotations

● To adjust size and style of the font:

[Link]("Frequency of Loan by Grade", fontsize=16, fontweight="bold", pad=15)

[Link]("Frequency", fontsize=14)

● Move annotation text:

[Link](text="Loans Grade E and below \nare very high risk.",
Position for the text xy=(4, 50),
How to move relative to point xytext=(-10, 30),
Didn’t set absolute value textcoords="offset points",...)
for text location

● Used LLM to:

● Create arrow
● Label each bar with
its frequency

Sean
SeanBarnes
Barnes
Data Visualization

Ticks & spines

Recap: Ticks & spines
● To rotate the x axis labels

[Link](rotation=0)

● To save the result of a plot method into a variable:

ax = sorted_grades.plot(kind="bar", color=colors)

● To add more ticks to plot:

[Link].set_minor_locator(AutoMinorLocator(2))

● To remove the spines from plot:

[Link]["left"].set_visible(False)

Sean
SeanBarnes
Barnes
 Data Visualization

Grouped column charts

Scenario 🏆 Goal: Identify key characteristics of loans from
states with highest average loan amount

● District of Columbia
● Alaska
● Hawaii

🎯 Task: Plot the loan amount by grade across these

three states using a grouped bar chart Alaska A

1. Set up your data:

Alaska B
● Filter data to only include top three states

You ● Group by state and grade Hawaii A

Data Analyst ● Select loan amount column and calculate ..
mean for each group .
● Create a grouped column chart showing the
mean value of loans of each grade

Sean
SeanBarnes
Barnes
 MultiIndex
Recap: Grouped column charts state grade
AK A 25750.000000
1. Selected the rows of interest B 30833.333333
C 12500.000000
# names of top three states for loan amount D 11100.000000
states = ["DC", "AK", "HI"] DC A 40000.000000
filtered_df = df[df["state"].isin(states)] B 12500.000000
C 25900.000000
HI A 1200.000000
2. Grouped by two features B 17733.333333
D 10000.000000
grouped_df = filtered_df.groupby(["state", "grade"]) E 31666.666667
F 28000.000000
3. Calculated mean of loan amount for each state Name: loan_amount, dtype: float64
and grade combination
✅ A lot of flexibility to create unique rows
grouped_loan_amount = grouped_df["loan_amount"].mean()
⛔ To plot, you’ll need to use .unstack()

Sean
SeanBarnes
Barnes
 .unstack()
state grade A B C D E F
AK A 25750.000000
B 30833.333333 AK 25750.0 30833.3 12500.0 11100.0 NaN NaN
C 12500.000000
D 11100.000000 DC 40000.0 12500.0 25900.0 NaN NaN NaN
DC A 40000.000000
B 12500.000000
HI 1200.0 17733.3 NaN 10000.0 31666.6 28000.0
C 25900.000000
HI A 1200.000000
B 17733.333333 grouped_loan_amount.unstack().plot(kind = "bar")
D 10000.000000
E 31666.666667
F 28000.000000 Columns are grouped by index automatically

Sean
SeanBarnes
Barnes
 Data Visualization

Stacked column charts

Scenario 🏆 Goal: Understand whether the composition of homeownership
changes based on the risk profile of the loan

🎯 Task: Highlight the proportion of renters for each group

Exploring total loan by:

● Grade
● Homeownership status

Perform similar steps:

● Grouping

You ● Aggregating
Data Analyst ● Unstacking before plotting

Sean
SeanBarnes
Barnes
Recap: Stacked column chart
● Create a stacked column chart rather than grouped one

grouped_df.plot(kind = "bar", stacked = True,

color=["lightgray", "darkgray", "mediumseagreen"])

● Compare within a category rather than across them

● Worked with LLM to develop 100% stacked bar chart:

proportion_df = grouped_df.div(grouped_df.sum(axis=1), axis=0)

proportion_df.plot(kind = "bar", stacked = True,

color=["lightgray", "darkgray", "mediumseagreen"])

● Easier comparison across the different categories

Sean
SeanBarnes
Barnes
Data Visualization

Scatter plots
Scenario 💡 Findings: Annual income is a moderate predictor of total
credit limit
○ Pearson correlation coefficient of 0.55

🎯 Tasks:

○ Graph how income

distribution impacts
amount of credit they
should offer
You
○ Visualize credit limits of
Data Analyst
incomes in top 5%

Sean
SeanBarnes
Barnes
Recap: Scatter plots Option Style

"^" ▲

● To create a scatter plot: "." ⬤

Columns

[Link](df["annual_income"], df["total_credit_limit"], alpha=0.5 , marker=".", color="darkgreen")

x-axis Transparency Style Color of marker

● To set x axis limit: [Link](0, 500000) [Link](alpha=0)

● To set y axis limit: [Link](...)

● To draw vertical line: [Link](x=top_5_percent_income, color="black", linestyle="--")

● To draw horizontal line: [Link](...)

Sean
SeanBarnes
Barnes
Data Visualization

Method chaining
Method chaining

● Process of linking several (((5+3)×2)−4)÷2

operations in a row

● Each operation depends

((8×2)−4)÷2
on the previous
( 16 − 4 ) ÷ 2
🔗 Analogy: Connecting
operations like links in a chain 12 ÷ 2
6

Sean
SeanBarnes
Barnes
 [Link](["grade", "homeownership"])["loan_amount"].sum().unstack().plot(kind="bar", stacked=True)

Select Sum loan Get data into Create

Group by grade & rows and
DataFrame loan_amount amounts for stacked
homeownership columns
column each group bar chart

amount
column
Groupby
from
Object
grouped
data

Sean
SeanBarnes
Barnes
 grouped_data = [Link](["grade", "homeownership"])

grouped_data_loan_amount = grouped_data["loan_amount"]

grouped_sum_of_loans = grouped_data_loan_amount.sum()

unstacked_sum_of_loans = grouped_sum_of_loans.unstack()

unstacked_sum_of_loans.plot(kind="bar", stacked=True)

Using variables for each step: Method chaining:

● Useful to save intermediate steps for later ● Get to plot as quickly as possible
● More flexible, but takes longer ● Less flexible
● If you wanted to stop at an intermediate
step, you wouldn’t be able to

Sean
SeanBarnes
Barnes
Data Visualization

Plotting with Seaborn

Seaborn
● A visualization tool

import seaborn

● Works well with matplotlib

● Main strengths:
○ Improved visual appeal
○ Reduced need to manipulate
the data
○ Additional plot types

Sean
SeanBarnes
Barnes
Plotting with Seaborn

At this point in analysis, you’re looking to: Seaborn can help in both:

⬜ Level up visual appeal of visualizations ✅ Improving visual appeal

⬜ Include many plots of different features ✅ Creating many charts quickly

and relationships in data
○ Including distributional charts

Sean
SeanBarnes
Barnes
Recap: Plotting with Seaborn
● Gives additional functions for plotting:
○ [Link]()
○ [Link]()
○ [Link]()

● Automatically summarizes your data: ● Set the estimation to:

Default
○ [Link]
[Link](selected_stocks, estimator = [Link])
○ [Link]
○ [Link]

Sean
SeanBarnes
Barnes
Data Visualization

Themes & palettes

Recap
● Use sns.set_theme() to change default styling

sns.set_theme(style = "white"
)

Default matplotlib look

● Gives more control of visual style of plots

● Used palette named argument

[Link](...palette = "Blues")

[Link](...palette ="RdYlGn_r")

● Avoid manually selecting colors

Sean
SeanBarnes
Barnes
Data Visualization

Box plots
Scenario
🏆 Goal: Characterizing different features present in the dataset

Features of each person: Features of individual loans:

● Occupation ● Loan amount

● Loan history ● Interest rate

You
Data Analyst

Sean
SeanBarnes
Barnes
 Horizontal ↔ Vertical ↕
Recap
Creates box plot: y = "interest_rate"
or

[Link](df, x = "interest_rate" )

Set both to segment by another variable:

[Link](df, x = "grade" , y = "interest_rate" )

Remove some axes from the plot:

[Link](left = True, bottom = True)

Increase the figure size:

[Link](figsize = (8, 6))
Width & height
in inches

Sean
SeanBarnes
Barnes
Data Visualization

Histograms
Scenario
🏆 Goal: Characterizing different features present in the dataset

Features of each person: Features of individual loans:

● Occupation ● Loan amount

● Loan history ● Interest rate

🎯 Task: Visualize using histogram to explain its unique properties

You
Data Analyst ● This visualization will help:
○ Identify and price common loan products

Sean
SeanBarnes
Barnes
Recap
Vertical bars Number of bins

To graph y = "interest_rate" bins = 10

histogram or or

[Link]( df, x = "grade" , binwidth = 50 , kde = True )

Horizontal bars Width of bins Density curve
estimate

sns.set_style("ticks" )

Sean
SeanBarnes
Barnes
Data Visualization

Other charts
Upcoming Plots

For each plot, you’ll see: ⛔ Try not to focus too much on code

Overview of the code ✅ Do keep in mind:

○ Commonalities with other
The output visualizations you’ve created
○ Output you expect from each
visualization

Sean
SeanBarnes
Barnes
Data Visualization

Combining charts
Matplotlib subplots
You can:

● Stack commands from seaborn and

matplotlib to create a chart

● Stack plots on top of each other

● Use technique to:

○ Stack complementary plots

○ Plot multiple distributions
together on the same axes

Sean
SeanBarnes
Barnes
Recap: Subplots
● Create a rugplot on histogram

● Disaggregated plot of individual

values

● Combine a stripplot with a boxplot

● Overlay histograms on the same

chart to compare distributions

● Accomplished by creating two

plots in the same figure

Sean
SeanBarnes
Barnes
Data Visualization

Matplotlib subplots
Scenario
🏆 Goal: Explore number of open credit lines
customers have
matplotlib Seaborn
○ Credit cards
○ Home equity +
○ Business line of credit

🎯 Task: Segmenting paid interest based on Create a grid of Fill up with

loan grade and open lines of credit empty subplots appealing plots

○ A lot of possible values for line of credit

○ Need to create a lot of graphs

Sean
SeanBarnes
Barnes
Recap: Matplotlib subplots
⬜ Create the figure
Canvas Width, Height
[Link](figsize = (5,5) ) 1 2 3

⬜ Use the matplolib .subplot()

Rows Col. Plot 4 5 6
[Link]( 2, 3, 4)

⬜ Create a plot

[Link](filtered_df, hue="grade", y="paid_interest", palette="RdYlGn_r")

Sean
SeanBarnes
Barnes
Data Visualization

Looping with subplots

Recap: Looping with subplots
Use for loop to iterate through subplots [Link](figsize=(15,15))

• Loop through a range: for i in range(1,10):

[Link](3, 3, i)
Starting: 1
filtered_df = df[df["open_credit_lines"] == i]
Ending: 1 after the last number
[Link]( filtered_df,
• Use to help improve code’s hue="grade",
efficiency and readability y="paid_interest",
palette="RdYlGn_r")
[Link](i)
[Link]("[Link]")

Sean
SeanBarnes
Barnes
Data Visualization

Seaborn pairplot
You’re almost done with EDA! 🎉
● Next steps: Create an appendix with many distributional plots

● Why: For clients interested in looking for their own insights

● How: Use Seaborn’s pairplot tool to create many plots quickly

Sean
SeanBarnes
Barnes
Recap: Seaborn pairplot
How to use [Link]() to:

● Create scatterplots between features and histogram for each feature individually

[Link](df[["loan_amount", "annual_income", "interest_rate", "paid_interest"]])

● Add to the title of the entire figure

[Link]("Pairplot of Loan Amount, Annual Income, Interest Rate, and Paid Interest")

Sean
SeanBarnes
Barnes