9  Lab: Seaborn

import warnings
warnings.filterwarnings('ignore')

In the previous notebooks we created basic, yet fast, visualisations using matplotlib. In this one we will be using a different plotting library called Seaborn. Some consider it the ggplot of Python with excellent default setting which make your data life easier.

9.1 Preparations

As usual, we need to load any package and data needed for our work.

import numpy as np
import pandas as pd
import seaborn as sns

office_df = pd.read_csv('data/raw/office_ratings.csv', encoding='UTF-8')

And check our data:

office_df.head()
season episode title imdb_rating total_votes air_date
0 1 1 Pilot 7.6 3706 2005-03-24
1 1 2 Diversity Day 8.3 3566 2005-03-29
2 1 3 Health Care 7.9 2983 2005-04-05
3 1 4 The Alliance 8.1 2886 2005-04-12
4 1 5 Basketball 8.4 3179 2005-04-19

We can try to replicate the same plots as in the previous notebooks.

9.2 Scatterplots

This is relatively similar to what we did in Section 7.3, but in this case we will be using seaborn’s replot() method.

sns.relplot(x='total_votes', y='imdb_rating', data=office_df)

9.2.1 Dates

If we want to create a scatterplot with dates, we will need to convert them to dates, too:

office_df['air_date'] =  pd.to_datetime(office_df['air_date'], errors='ignore')

g = sns.relplot(x="air_date", y="imdb_rating", kind="scatter", data=office_df)

9.3 Functions

We can define our own functions. A function helps us with code we are going to run multiple times. For instance, the below function scales values between 0 and 1.

Here is a modified function from stackoverflow.

office_df.head()
season episode title imdb_rating total_votes air_date
0 1 1 Pilot 7.6 3706 2005-03-24
1 1 2 Diversity Day 8.3 3566 2005-03-29
2 1 3 Health Care 7.9 2983 2005-04-05
3 1 4 The Alliance 8.1 2886 2005-04-12
4 1 5 Basketball 8.4 3179 2005-04-19 
def normalize(df, feature_name):
    result = df.copy()
    
    max_value = df[feature_name].max()
    min_value = df[feature_name].min()
    
    result[feature_name] = (df[feature_name] - min_value) / (max_value - min_value)
    
    return result

Passing the dataframe and name of the column will return a dataframe with that column scaled between 0 and 1.

normalize(office_df, 'imdb_rating')
season episode title imdb_rating total_votes air_date
0 1 1 Pilot 0.300000 3706 2005-03-24
1 1 2 Diversity Day 0.533333 3566 2005-03-29
2 1 3 Health Care 0.400000 2983 2005-04-05
3 1 4 The Alliance 0.466667 2886 2005-04-12
4 1 5 Basketball 0.566667 3179 2005-04-19
... ... ... ... ... ... ...
183 9 19 Stairmageddon 0.433333 1484 2013-04-11
184 9 20 Paper Airplane 0.433333 1482 2013-04-25
185 9 21 Livin' the Dream 0.733333 2041 2013-05-02
186 9 22 A.A.R.M. 0.866667 2860 2013-05-09
187 9 23 Finale 1.000000 7934 2013-05-16

188 rows × 6 columns

Replacing the origonal dataframe. We can normalize both out votes and rating.

office_df = normalize(office_df, 'imdb_rating')
office_df = normalize(office_df, 'total_votes')
office_df
season episode title imdb_rating total_votes air_date
0 1 1 Pilot 0.300000 0.353616 2005-03-24
1 1 2 Diversity Day 0.533333 0.332212 2005-03-29
2 1 3 Health Care 0.400000 0.243082 2005-04-05
3 1 4 The Alliance 0.466667 0.228253 2005-04-12
4 1 5 Basketball 0.566667 0.273047 2005-04-19
... ... ... ... ... ... ...
183 9 19 Stairmageddon 0.433333 0.013912 2013-04-11
184 9 20 Paper Airplane 0.433333 0.013606 2013-04-25
185 9 21 Livin' the Dream 0.733333 0.099067 2013-05-02
186 9 22 A.A.R.M. 0.866667 0.224278 2013-05-09
187 9 23 Finale 1.000000 1.000000 2013-05-16

188 rows × 6 columns

9.3.1 Long format

Seaborn prefers a long format table. Details of melt can be found here.

office_df_long=pd.melt(office_df, id_vars=['season', 'episode', 'title', 'air_date'], value_vars=['imdb_rating', 'total_votes'])
office_df_long
season episode title air_date variable value
0 1 1 Pilot 2005-03-24 imdb_rating 0.300000
1 1 2 Diversity Day 2005-03-29 imdb_rating 0.533333
2 1 3 Health Care 2005-04-05 imdb_rating 0.400000
3 1 4 The Alliance 2005-04-12 imdb_rating 0.466667
4 1 5 Basketball 2005-04-19 imdb_rating 0.566667
... ... ... ... ... ... ...
371 9 19 Stairmageddon 2013-04-11 total_votes 0.013912
372 9 20 Paper Airplane 2013-04-25 total_votes 0.013606
373 9 21 Livin' the Dream 2013-05-02 total_votes 0.099067
374 9 22 A.A.R.M. 2013-05-09 total_votes 0.224278
375 9 23 Finale 2013-05-16 total_votes 1.000000

376 rows × 6 columns

Which we can plot in seaborn like so.

sns.relplot(x='air_date', y='value', size='variable', data=office_df_long)

?sns.relplot