How to create a correlation heatmap in Python?

Building a Correlation Heatmap in Python Using Seaborn

Correlation heatmaps are a powerful visual tool to examine relationships between variables. By computing the correlation matrix and mapping values to colors, Seaborn’s heatmap() makes it easy to identify strong associations, trends, and potential multicollinearity in your dataset.

Why Use a Correlation Heatmap?

A correlation heatmap reveals the degree to which variables move together—whether positively, negatively, or not at all. This insight is crucial for feature selection, multivariate analysis, and model building. Color gradients allow you to spot patterns at a glance.

Step 1: Compute the Correlation Matrix

First, calculate pairwise correlation coefficients using pandas:

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

# Load a sample dataset
df = sns.load_dataset("iris")

# Compute correlation matrix (ignores non-numeric)
corr = df.corr()
corr

The resulting matrix shows correlation values between each numeric variable pair.

Step 2: Draw the Heatmap

Use Seaborn’s heatmap() to display the matrix:

sns.heatmap(corr, annot=True, fmt=".2f", cmap="coolwarm")
plt.title("Correlation Heatmap for Iris Data")
plt.show()

Here, annot=True prints values in cells, and cmap="coolwarm" provides a diverging palette for visual balance.

Step 3: Format and Style the Plot

• vmin/vmax: Standardize color range across heatmaps.
• linewidths: Add cell separation lines.
• fmt: Control numeric formatting (e.g., two decimals).

sns.heatmap(corr, annot=True, fmt=".2f", cmap="RdBu", linewidths=0.5,
            vmin=-1, vmax=1)
plt.title("Styled Correlation Heatmap")
plt.show()

Step 4: Mask Redundant Areas

Optionally hide the upper triangle to reduce redundancy:

mask = np.triu(np.ones_like(corr, dtype=bool))
sns.heatmap(corr, mask=mask, annot=True, cmap="coolwarm",
            linewidths=0.5, fmt=".2f")
plt.title("Correlation Heatmap (Lower Triangle Only)")
plt.show()

This highlights only unique variable pairs on the plot.

Step 5: Interpret the Heatmap

Look for:

• Dark reds/blues: strong positive or negative correlations.
• Near-zero colors: weak or no relationship.
• Pairs with high correlation—useful for identifying variable redundancy.

Color-coded overlays allow fast and informed evaluation of relationships.

Bonus: Heatmap for Non-Numeric Variables

Convert categorical data into numeric form for correlation, e.g., using one-hot encoding. Then, apply the same visualization steps.

df_enc = pd.get_dummies(df, drop_first=True)
corr_enc = df_enc.corr()
sns.heatmap(corr_enc, annot=False, cmap="viridis")
plt.title("Heatmap with Encoded Variables")
plt.show()

Conclusion

Seaborn’s heatmap() makes creating correlation visuals in Python both simple and expressive. With options to mask, style, and annotate, you can tailor heatmaps to your analytical needs—making them ideal for exploring dataset structure and guiding modeling efforts.