5 Best Ways to Plot a Heatmap for 3 Columns in Python with Seaborn

by

💡 Problem Formulation: Visualizing relationships across multiple variables in a dataset can be challenging. For data analysts and scientists using Python, a common approach might be to create a heatmap which communicates the correlations or interactions between the variables effectively. For example, given a dataset with columns ‘A’, ‘B’, and ‘C’, the desired output would be a clear heatmap that illustrates the interconnectedness of these three columns.

Method 1: Basic Heatmap Using Seaborn’s heatmap Function

This method involves using Seaborn’s heatmap() function to transform a correlation matrix generated from the three columns into a visual heatmap. The function provides a high-level interface to draw attractive and informative heatmaps with extensive customization options.

Here’s an example:

import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd

# Sample DataFrame
data = pd.DataFrame({
    'A': [34, 79, 56],
    'B': [22, 45, 33],
    'C': [59, 85, 55]
})

# Calculating the correlation
correlation_matrix = data.corr()

# Plotting the heatmap
sns.heatmap(correlation_matrix, annot=True)
plt.show()

The output is a graphical heatmap with shaded squares representing the strength of correlation between columns ‘A’, ‘B’, and ‘C’.

This snippet first calculates the correlation matrix of the three columns using Pandas’ corr() method. The sns.heatmap() function then takes this matrix to create the heatmap, with the annot=True parameter adding the actual correlation values as text on the heatmap for better clarity.

Method 2: Heatmap with a Mask for Upper Triangle

To enhance readability, heatmaps can display only the lower triangle of the correlation matrix since it is symmetric. Seaborn and Matplotlib offer simple ways to create a mask that hides the upper triangle.

Here’s an example:

import seaborn as sns
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd

# Sample DataFrame
data = pd.DataFrame({
    'A': [34, 79, 56],
    'B': [22, 45, 33],
    'C': [59, 85, 55]
})

# Calculating the correlation
correlation_matrix = data.corr()

# Create a mask to hide the upper triangle
mask = np.triu(np.ones_like(correlation_matrix, dtype=bool))

# Plotting the heatmap with the mask
sns.heatmap(correlation_matrix, annot=True, mask=mask)
plt.show()

The output is a heatmap with only the lower triangle showing the correlations between columns ‘A’, ‘B’, and ‘C’.

This method improves upon the first by adding a mask with np.triu() (triangular upper) to create an array that will hide the upper triangle when passed to the sns.heatmap() function, ensuring that each correlation is displayed only once.

Method 3: Customizing the Color Palette

Customizing the color palette of the heatmap allows for better visual appeal and can make it easier to read the heatmap. Seaborn offers a range of color palettes which can be applied easily.

Here’s an example:

import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd

# Sample DataFrame
data = pd.DataFrame({
    'A': [34, 79, 56],
    'B': [22, 45, 33],
    'C': [59, 85, 55]
})

# Calculating the correlation
correlation_matrix = data.corr()

# Plotting the heatmap with a coolwarm palette
sns.heatmap(correlation_matrix, annot=True, cmap='coolwarm')
plt.show()

The output is a colorful heatmap representing the correlations between ‘A’, ‘B’, and ‘C’ with a blue to red gradient.

This code introduces the cmap argument to the heatmap() function, specifying the ‘coolwarm’ palette that provides a visually pleasing gradient from cool to warm colors, effectively highlighting the differences in correlation values.

Method 4: Heatmap with Hierarchical Clustering

Seaborn’s clustermap() function can be utilized for plotting a heatmap that also includes hierarchical clustering performed along the rows and columns. This can reveal patterns reflecting the underlying structure of the data.

Here’s an example:

import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd

# Sample DataFrame
data = pd.DataFrame({
    'A': [34, 79, 56],
    'B': [22, 45, 33],
    'C': [59, 85, 55]
})

# Plotting the heatmap with clustering
sns.clustermap(data.corr(), annot=True)
plt.show()

The output is a heatmap with dendrograms added to both axes, showing how the columns cluster together.

This code calls the clustermap() function on the correlation matrix, resulting in a heatmap that clusters the rows and columns based on their similarity. The dendrograms provide insights into the clustering process and can help identify groupings within the variables.

Bonus One-Liner Method 5: Inverting the Heatmap

Inverting the heatmap’s color scale can sometimes highlight negative correlations more effectively, which could be beneficial depending on the data’s context.

Here’s an example:

sns.heatmap(data.corr(), annot=True, center=0, cmap='vlag')

The output is a heatmap where the center point is zero, emphasizing both positive and negative correlations equally.

By setting the center parameter to 0 and choosing the ‘vlag’ colormap, this code snippet creates an inverted color scale heatmap that makes it easier to distinguish between negative and positive correlations visually.

Summary/Discussion

  • Method 1: Basic Heatmap. Simple implementation. Limited visual output.
  • Method 2: Heatmap with Mask. More readable by avoiding redundancy. Requires extra step for mask creation.
  • Method 3: Custom Color Palette. Enhances visual appeal. Color choice might not suit all datasets.
  • Method 4: Heatmap with Clustering. Provides additional structure information. Potentially more complex interpretation.
  • Method 5: Inverted Heatmap. Highlights negative correlations. May miss out on simplicity of traditional heatmap coloring.