seaborn

Seaborn

Seaborn is a high-level wrapper around Matplotlib. It makes statistical plots (violins, heatmaps, pairs) easy.

When to Use

• Statistical Data Visualization: Producing clear distribution plots, regression curves, and correlation heatmaps.
• Multi-Plot Categorical Comparisons: Visualizing complex distributions with boxplot, violinplot, and catplot.
• Modern Object Interface (sns.objects): Clean grammar of graphics visualization inspired by ggplot2.
• Pairwise Feature Exploration: Rapidly identifying variable correlations and cluster separations with pairplot.

Quick Start

import seaborn as sns
import matplotlib.pyplot as plt

# Load sample dataset and plot regression with confidence interval
tips = sns.load_dataset("tips")

plt.figure(figsize=(8, 5)) sns.set_theme(style="whitegrid") sns.scatterplot(data=tips, x="total_bill", y="tip", hue="time", size="size") plt.title("Tip Amount vs Total Bill by Dining Time") plt.show()

## Core Concepts

### Modern Seaborn Objects Interface (sns.objects)

Grammar of graphics data visualization:

```python
import seaborn.objects as so
import seaborn as sns
import matplotlib.pyplot as plt

# Load sample dataset
penguins = sns.load_dataset("penguins").dropna()

# Declarative grammar of graphics plot
plot = (
    so.Plot(penguins, x="bill_length_mm", y="bill_depth_mm", color="species", pointsize="body_mass_g")
    .add(so.Dots(alpha=0.7))
    .add(so.Line(), so.PolyFit(order=1))
    .label(
        title="Penguin Bill Dimensions & Linear Fits",
        x="Bill Length (mm)",
        y="Bill Depth (mm)",
        color="Species"
    )
    .theme({"axes.grid": True})
)

fig = plt.figure(figsize=(9, 5))
plot.on(fig).save("seaborn_objects_plot.png")

Statistical Distributions & Faceted Grids

Visualizing density and distributions across categories:

import seaborn as sns
import matplotlib.pyplot as plt

sns.set_theme(style="whitegrid", palette="muted")
tips = sns.load_dataset("tips")

# Multi-panel faceted distribution
g = sns.displot(
    data=tips,
    x="total_bill",
    hue="time",
    col="day",
    kind="kde",
    fill=True,
    common_norm=False,
    palette="crest",
    height=3.5,
    aspect=0.8
)

g.set_axis_labels("Total Bill ($)", "Density")
g.savefig("tips_kde_faceted.png", bbox_inches="tight")
plt.close()

Categorical Box and Violin Comparisons

Comparing distributions across discrete classes:

fig, ax = plt.subplots(figsize=(8, 5))

sns.violinplot(
    data=tips,
    x="day",
    y="total_bill",
    hue="sex",
    split=True,
    inner="quart",
    palette={"Male": "#3b82f6", "Female": "#ec4899"},
    ax=ax
)

ax.set_title("Total Bill Distribution by Day & Gender")
plt.savefig("violin_comparison.png", bbox_inches="tight")
plt.close(fig)

Common Patterns

Correlation Heatmap with Masked Triangle

Problem: Large correlation heatmaps are visually redundant and cluttered when showing both upper and lower triangles.

Solution: Mask upper triangle using NumPy:

import numpy as np

corr = df.select_dtypes(include=np.number).corr()
mask = np.triu(np.ones_like(corr, dtype=bool))

plt.figure(figsize=(10, 8))
sns.heatmap(corr, mask=mask, cmap="vlag", vmax=1.0, vmin=-1.0,
            annot=True, fmt=".2f", square=True, linewidths=.5)
plt.title("Feature Correlation Matrix")
plt.show()

Best Practices

Do:

• Adopt the new seaborn.objects (so.Plot) API for modern, composable grammar-of-graphics visualizations.
• Apply sns.set_theme() at application startup to configure consistent typography and aesthetic palettes.
• Pass tidy long-form DataFrames to Seaborn functions (data=df, x='col1', y='col2', hue='category').
• Always close Matplotlib figures (plt.close(fig)) when generating charts in backend web pipelines.

Don't:

• Use pie charts for categorical proportions; use horizontal bar charts (sns.barplot).
• Overload charts with more than 4-5 categories in hue; use faceted subplots (col='category') instead.
• Mix stateful plt.title() with object-oriented ax.set_title().

Troubleshooting

Error	Cause	Solution
TypeError: Object of type '...' is not JSON serializable	Passing non-numeric columns to heatmap().	Filter numeric columns: df.select_dtypes(include=np.number).
Seaborn plot showing tiny unreadable fonts	Default scaling mismatch with high-DPI displays.	Set context scaling: sns.set_context("talk") or "poster".
AttributeError: module 'seaborn' has no attribute '...'	Outdated Seaborn version installed.	Upgrade Seaborn to 0.13+: pip install --upgrade seaborn.

References

• Seaborn Documentation