Python Seaborn Tutorial for Beginners
What is Seaborn?
• Python Data Visualization Library - based on MatPlotLib (see previous tutorials)
• Used for plotting statistical graphs, identifying trends, relationships & outliers
• In my opinion, Seaborn is easier & faster to use (less code) Vs MatPlotLib
# Visual representation of Seaborn
import os
from IPython.display import Image
PATH = "F:\\Github\\Python tutorials\\Introduction to Seaborn\\"
Image(filename = PATH + "Seaborn.png", width=900, height=900)
Tutorial Overview
• What is matplotlib and how/why it's used
• Trend Plots:
– Line Plots
• Summary Plots:
– Bar Plots
• Distribution of Data:
– Histogram
– Box Plots
• Relationship Plots
– Scatter Plots
– lmplot (combo of regplot() and FacetGrid)
• Holistic views / Combo:
– Sub Plots
– Pair Plots
– Join Plots
• Correlation / Relationships:
– Heat Maps
Tutorial Overview by video
Video 1:
1. What is matplotlib and how/why it's used
2. Line Plots
3. Bar Plots
4. Histogram
Video 2:
1. Box Plots
2. Scatter Plots
3. lmplot (combo of regplot() and FacetGrid)
Video 3:
1. Sub Plots
2. Pair Plots
3. Join Plots
4. Heat Maps
Importing / Installing packages
# Packages / libraries
import os #provides functions for interacting with the operating
system
import numpy as np
import pandas as pd
from matplotlib import pyplot as plt
import matplotlib.pyplot as plt
import seaborn as sns
%matplotlib inline
# To install seaborn type "pip install seaborn" to the anaconda
terminal
# import sys
# !conda list Check the packages installed
1. Loading the Raw Data
# Loading the data
raw_data = pd.read_csv("F:\\Github\Python tutorials\\Introduction to
Seaborn\\Marketing Raw Data.csv")
# runs all the data
raw_data
print(raw_data.shape)
#runs the first 5 rows
raw_data.head()
(182, 11)
Date Week Week_ID Month Month_ID Year Day_Name
Visitors \
0 09/11/2020 46 34 11 11 2020 Monday
707
1 10/11/2020 46 34 11 11 2020 Tuesday
1455
2 11/11/2020 46 34 11 11 2020 Wednesday
1520
3 12/11/2020 46 34 11 11 2020 Thursday
1726
4 13/11/2020 46 34 11 11 2020 Friday
2134
Revenue Marketing Spend Promo
0 465 651.375 No Promo
1 10386 1298.250 Promotion Red
2 12475 1559.375 Promotion Blue
3 11712 1801.750 No Promo
4 10000 2614.500 No Promo
2. Line Gragh
# Example 1 - Simple 1 line graph
# Assuming we want to investigate the Revenue by Date
ax = sns.lineplot(x='Week_ID', y='Revenue', data = raw_data)
# Notes: error bands show the confidence interval
# Example 2 - Adding Categories
# By Promo
ax = sns.lineplot(x='Week_ID', y='Revenue', hue = 'Promo', data =
raw_data)
# Example 3 - By Promo with style
ax = sns.lineplot(x='Week_ID', y='Revenue', hue = 'Promo', style =
'Promo', data = raw_data)
# Example 4 - By Promo with style & Increase the size & Remove error
bars
# increase the size
sns.set(rc={'figure.figsize':(12,10)})
ax = sns.lineplot(x='Week_ID', y='Revenue', hue = 'Promo', style =
'Promo', data = raw_data, ci=None)
# Example 5 - By Promo with style & Increase the size & Remove error
bars & adding markers & by month
ax = sns.lineplot(x='Month_ID', y='Revenue', hue = 'Promo', style =
'Promo', data = raw_data, ci=None, markers=True)
# Example 6 - By Day_Name with style & Increase the size & Remove
error bars & adding markers & by month
ax = sns.lineplot(x='Month_ID', y='Revenue', hue = 'Day_Name', style =
'Promo', data = raw_data, ci=None, markers=True)
# Example 7 - By Year with style & Increase the size & Remove error
bars & adding markers & by month
ax = sns.lineplot(x='Year', y='Revenue', hue = 'Day_Name', style =
'Promo', data = raw_data, ci=None, markers=True)
LinePlot Documentation
https://seaborn.pydata.org/generated/seaborn.lineplot.html
3. Bar Plots
# Example 1 - Total Revenue by Month
ax = sns.barplot(x="Month_ID", y="Revenue", data=raw_data)
# Notes:
# 1 - the lines signify the confidence interval
# 2 - Takes mean by default
raw_data[['Month_ID', 'Revenue']].groupby('Month_ID', as_index =
False).agg({'Revenue':'mean'})
Month_ID Revenue
0 11 11255.454545
1 12 11667.806452
2 13 9588.516129
3 14 10683.892857
4 15 10555.354839
5 16 10806.500000
6 17 7636.000000
# Example 2 - Total Revenue by Month - Remove the Confidence Interval
ax = sns.barplot(x="Month_ID", y="Revenue", data=raw_data, ci=False)
# Example 3 - Total Revenue by Month - Remove the Confidence Interval
- By Promo
ax = sns.barplot(x="Month_ID", y="Revenue", data=raw_data, ci=False,
hue = 'Promo')
# Example 4 - Total Revenue by Month - Remove the Confidence Interval
- By Promo - Changing direction
ax = sns.barplot(x="Revenue", y="Year", ci=False, hue = 'Promo',
orient = 'h', data=raw_data)
# Example 5 - Total Revenue by Month - Remove the Confidence Interval
- By Promo - Changing direction - Changing Colour
ax = sns.barplot(x="Revenue", y="Year", ci=False, hue = 'Promo',
orient = 'h', data=raw_data, color="#1CB3B1")
# Cool Way to pick colours
# https://htmlcolorcodes.com/color-picker/
4. Histograms
# Example 1 - Investigating the distribution of Revenue
x = raw_data['Revenue'].values
sns.distplot(x, color = 'blue');
# As you can see, it's a bit imbalance. Right skewd distribution as
the mean is to the right
# Example 2 - Investigating the distribution of Revenue, adding the
mean
x = raw_data['Revenue'].values
sns.distplot(x, color = 'blue');
# Calculating the mean
mean = raw_data['Revenue'].mean()
#ploting the mean
plt.axvline(mean, 0,1, color = 'red')
<matplotlib.lines.Line2D at 0x1f6ca13ca58>
# Example 3 - Investigating the distribution of Visitors, adding the
mean
x = raw_data['Visitors'].values
sns.distplot(x, color = 'red');
# Calculating the mean
mean = raw_data['Visitors'].mean()
#ploting the mean
plt.axvline(mean, 0,1, color = 'blue')
<matplotlib.lines.Line2D at 0x1f6c9ba4358>
5. Box Plots
# Example 1 - Investigating the distribution of Revenue
x = raw_data['Revenue'].values
ax = sns.boxplot(x)
print('The meadian is: ', raw_data['Revenue'].median())
# Notes:
# The line signifies the median
# The box in the middle show the beginning of Q1 (25th percentile) and
the end of the Q3 (75th percentile)
# The whiskers (left - right) show the minimum quartile and maximum
quartile
# The dots on the right are "outliers"
The meadian is: 9452.0
# More Details
PATH = "F:\\Github\\Python tutorials\\Introduction to Seaborn\\"
Image(filename = PATH + "Seaborn boxplot.png", width=900, height=900)
# More details here: https://towardsdatascience.com/understanding-
boxplots-5e2df7bcbd51
# Credits: Michael Galarnyk
# Example 2 - Investigating the distribution of Revenue by Day
ax = sns.boxplot(x="Day_Name", y="Revenue", data=raw_data)
# Example 3 - Investigating the distribution of Revenue by Day -
Horizontal - change color
ax = sns.boxplot(x="Revenue", y="Day_Name", data=raw_data, color =
'#EE67CF')
# Cool Way to pick colours
# https://htmlcolorcodes.com/color-picker/
# Example 4 - Investigating the distribution of Revenue by Day -
changing color - adding hue
ax = sns.boxplot(x="Day_Name", y="Revenue", data=raw_data,
color="#B971E7", hue = 'Promo')
# Cool Way to pick colours
# https://htmlcolorcodes.com/color-picker/
# Example 5 - Investigating the distribution of Revenue by Day - by
color - by data points
ax = sns.boxplot(x="Day_Name", y="Revenue", data=raw_data, color =
'#D1EC46')
ax = sns.swarmplot(x="Day_Name", y="Revenue", data=raw_data,
color="red")
More on Boxplots here:
https://seaborn.pydata.org/generated/seaborn.boxplot.html#seaborn.boxplot
6. ScatterPlots
raw_data.columns
Index(['Date', 'Week', 'Week_ID', 'Month', 'Month_ID', 'Year',
'Day_Name',
'Visitors', 'Revenue', 'Marketing Spend', 'Promo'],
dtype='object')
# Example 1 - Relationship between Marketing Spend and Revenue
ax = sns.scatterplot(x="Marketing Spend", y="Revenue", data=raw_data,
color = 'green')
# Example 2 - Relationship between Marketing Spend and Revenue -
changing color, hue & Style
ax = sns.scatterplot(x="Marketing Spend", y="Revenue", data=raw_data,
color = 'green', hue = 'Promo', style = 'Promo')
# Example 3 - Relationship between Marketing Spend and Revenue -
changing color & hue - adding size
ax = sns.scatterplot(x="Marketing Spend", y="Revenue", data=raw_data,
color = 'green', hue = 'Promo', size = 'Revenue',
sizes=(20, 200))
# Example 4 - Relationship between Marketing Spend and Revenue -
changing color & hue - adding size - by day
ax = sns.scatterplot(x="Visitors", y="Revenue", data=raw_data, color =
'green', hue = 'Day_Name', size = 'Revenue',
sizes=(20, 200))
7. lmPlots
# Example 1 - Relationship between Marketing Spend and Revenue
ax = sns.lmplot(x="Marketing Spend", y="Revenue", data=raw_data,
height=9)
# Notes:
# What is Linear Regression: It is a predictive statistical method for
modelling the relationship between x (independent variable) & y
(dependent V).
# How it works (cost function MSE):
https://towardsdatascience.com/machine-learning-fundamentals-via-
linear-regression-41a5d11f5220
# Example 2 - Relationship between Marketing Spend and Revenue -
changing color, hue & Style
ax = sns.lmplot(x="Marketing Spend", y="Revenue", data=raw_data, hue =
'Promo', ci= False, height=9, markers=["o", "x", "+"])
# Example 3 - Relationship between Marketing Spend and Revenue - by
column
ax = sns.lmplot(x="Marketing Spend", y="Revenue", data=raw_data, col =
'Promo', ci= False, height=5,
line_kws={'color': 'red'},
scatter_kws={'color':'#ADC11E'})
# Example 4 - Relationship between Marketing Spend and Revenue - by
column - by day - add Jitter too
ax = sns.lmplot(x="Visitors", y="Revenue", data=raw_data, col =
'Day_Name', ci= False, height=4,
line_kws={'color': '#031722'},
scatter_kws={'color':'#1E84C1'},
col_wrap=3,
x_jitter=.3)
8. SubPlots
# Set up the matplotlib figure
fig, axes = plt.subplots(2, 2, figsize=(12, 7))
a = raw_data['Revenue'].values
b = raw_data['Visitors'].values
c = raw_data['Marketing Spend'].values
# plot 1
sns.distplot(a, color = 'blue', ax=axes[0,0])
# plot 2
sns.distplot(b, color = 'blue', ax=axes[0,1])
# plot 3
sns.distplot(c, color = 'blue', ax=axes[1,0])
# plot 4
sns.boxplot(x="Revenue", y="Day_Name", data=raw_data, color =
'#52F954', ax=axes[1,1])
sns.swarmplot(x="Revenue", y="Day_Name", data=raw_data, color="blue",
ax=axes[1,1])
<matplotlib.axes._subplots.AxesSubplot at 0x1f6cde977f0>
9. Pairplots
# Example 1 - running on all dataframe - green color
g = sns.pairplot(raw_data, plot_kws={'color':'green'})
# Example 2 - running on specific columns - green color
g = sns.pairplot(raw_data[['Revenue','Visitors','Marketing Spend']],
plot_kws={'color':'#0EDCA9'})
# Cool Way to pick colours
# https://htmlcolorcodes.com/color-picker/
# Example 3 - running on specific columns - adding hue
g = sns.pairplot(raw_data[['Revenue','Visitors','Marketing Spend',
'Promo']], hue = 'Promo')
# Example 4 - running on specific columns - adding hue - adding kind =
reg
g = sns.pairplot(raw_data[['Revenue','Visitors','Marketing Spend',
'Promo']], hue = 'Promo', kind="reg", size = 6)
More on Pairplots:
https://seaborn.pydata.org/generated/seaborn.pairplot.html
10. JoinPlots
Draw a plot of two variables with bivariate and univariate graphs.
# Example 1 - Revenue vs marketing Spend Relationship with
g = sns.jointplot("Revenue", "Marketing Spend", data=raw_data,
kind="reg", color = 'green', size = 10)
11. Heat Map
# First we need to create a "Dataset" to display on a Heatmap - we
will use a correlation dataset
# .corr() is used to find the pairwise correlation of all columns in
the dataframe. Any null values are automatically excluded
# The closer to 1 or -1 the better. As one variable increases, the
other variable tends to also increase / decrease
# More Info here: https://statisticsbyjim.com/basics/correlations/
# Example 1 - Heatmap for PC
pc = raw_data[['Revenue','Visitors','Marketing Spend',
'Promo']].corr(method ='pearson')
cols = ['Revenue CC','Visitors CC','Marketing Spend CC']
ax = sns.heatmap(pc, annot=True,
yticklabels=cols,
xticklabels=cols,
annot_kws={'size': 50})
# Example 2 - Heatmap for PC - changing cmap
ax = sns.heatmap(pc, annot=True,
yticklabels=cols,
xticklabels=cols,
annot_kws={'size': 50},
cmap="BuPu")
# Examples:
# cmap="YlGnBu"
# cmap="Blues"
# cmap="BuPu"
# cmap="Greens"
More details for Heatmaps here:
https://seaborn.pydata.org/generated/seaborn.heatmap.html
