Heatmap
A heatmap is a method for visualizing two-dimensional data using colors to represent the magnitude of each value.
This example uses a heatmap to represent pairwise correlations between variables in a dataset.
Creating a Simple Heatmap
seaborn
has a function called heatmap()
. Its only required parameter is data
which should be a 2D (rectangle) dataset.
Perhaps the most common use case of a heatmap is with a correlation matrix like in the example above. Given a DataFrame
, we should first call its corr()
method to get a correlation matrix and only then pass this matrix as an argument for the heatmap()
function:
A common use case for a heatmap is displaying a correlation matrix. Given a DataFrame
, first call its corr()
method to obtain the correlation matrix, then pass this matrix as an argument to the heatmap()
function.
import seaborn as sns import matplotlib.pyplot as plt import pandas as pd # Loading the dataset with the countries data url = 'https://content-media-cdn.codefinity.com/courses/47339f29-4722-4e72-a0d4-6112c70ff738/countries_data.csv' countries_df = pd.read_csv(url, index_col=0) # Creating a correlation matrix with all numeric variables correlation_matrix = countries_df.corr(numeric_only=True) # Creating a heatmap based on the correlation matrix sns.heatmap(correlation_matrix) plt.show()
The correlation matrix was created using only the numeric columns of the DataFrame
. Columns containing strings were excluded by setting numeric_only=True
.
Annotation and Colors
This heatmap can be made more informative via writing the appropriate value (correlation coefficient in our case) in each cell . That can be done simply by setting the annot
parameter to True
.
It is also possible to change the colors for our heatmap via setting the cmap
parameter (you can explore it in the "Choosing color palettes" article).
import seaborn as sns import matplotlib.pyplot as plt import pandas as pd # Loading the dataset with the countries data url = 'https://content-media-cdn.codefinity.com/courses/47339f29-4722-4e72-a0d4-6112c70ff738/countries_data.csv' countries_df = pd.read_csv(url, index_col=0) # Creating a correlation matrix with all numeric variables correlation_matrix = countries_df.corr(numeric_only=True) # Setting annotation and color palette sns.heatmap(correlation_matrix, annot=True, cmap='viridis') plt.show()
The color bar on the right can be removed by setting cbar=False
.
In most of the cases that's all you will need from a heatmap customization, however, you can always explore more in heatmap()
documentation.
Improving Readability
The final thing that would improve the readability of our heatmap is rotating the ticks using already familiar xticks()
and yticks()
functions:
import seaborn as sns import matplotlib.pyplot as plt import pandas as pd # Loading the dataset with the countries data url = 'https://content-media-cdn.codefinity.com/courses/47339f29-4722-4e72-a0d4-6112c70ff738/countries_data.csv' countries_df = pd.read_csv(url, index_col=0) # Creating a correlation matrix with all numeric variables correlation_matrix = countries_df.corr(numeric_only=True) # Creating a heatmap based on the correlation matrix sns.heatmap(correlation_matrix, annot=True, cmap='viridis') # Rotating the ticks by 20 degrees counterclockwise plt.xticks(rotation=20) plt.yticks(rotation=20) plt.show()
Swipe to start coding
- Use the correct method to create a correlation matrix.
- Set the argument of the method to include only numeric variables.
- Use the correct function to create a heatmap.
- Set
correlation_matrix
to be the data for the heatmap via specifying the first argument. - Add the values in each cell of the matrix via specifying the second argument.
- Set the palette (color map) of the heatmap to
'crest'
via specifying the third (rightmost) argument. - Rotate x-axis and y-axis ticks by 15 degrees counterclockwise via specifying a keyword argument in
xticks()
andyticks()
.
Løsning
Takk for tilbakemeldingene dine!