Heatmaps & Correlations

Purpose

Heatmaps visualize two-dimensional data (matrices) using color intensity. Common uses:

• Correlation matrices (how variables relate).
• Confusion matrices (classification performance).
• Time-series aggregations (e.g., activity by hour and day).

Example in the Notebook

The notebook computes and displays the correlation matrix of iris features:

• Sepal length, sepal width, petal length, petal width.
• Uses a diverging color palette (coolwarm) centered at 0.
• Annotations show exact correlation values.

Key Code Snippet

corr = iris.select_dtypes(include=np.number).corr()
sns.heatmap(corr, annot=True, cmap='coolwarm', center=0, ax=ax)

Customization Tips

• Color map: Use cmap='viridis' (perceptually uniform), cmap='RdBu_r' (diverging), or cmap='Greens' (sequential).
• Annotation: Set annot=True to show values, or annot=False to hide.
• Centering: Use center=0 for diverging colormaps to anchor at a meaningful value.
• Normalization: Use vmin and vmax to control the color scale range.
• Clustermap: Use sns.clustermap() to hierarchically cluster rows/columns.

When to Use

• Always: for visualizing correlations and matrices.
• Consider: clustering when you want to identify related variables/samples.
• Avoid: heatmaps for more than ~20×20 cells without annotation clarity.

See Also

• Seaborn: heatmap(), clustermap()
• Matplotlib: imshow()