Seaborn: Statistical Visualization

Matplotlib can draw anything, but common statistical charts take a lot of boilerplate. Seaborn sits on top of matplotlib and specializes in exactly those charts: it understands DataFrames directly, maps columns to colors and styles for you, and looks great out of the box.

What seaborn adds

Compare drawing a scatter plot colored by category. In matplotlib you filter the DataFrame per group and call scatter in a loop; in seaborn you just name the columns:

import seaborn as sns
import matplotlib.pyplot as plt
 
tips = sns.load_dataset("tips")   # a classic demo dataset: restaurant bills
tips.head()

# one line: x, y, and "color by smoker status"
sns.scatterplot(data=tips, x="total_bill", y="tip", hue="smoker")
plt.show()

That data= + column-names interface is seaborn's core idea. The three semantic mappings you'll use constantly:

• hue — map a column to color
• style — map a column to marker shape
• size — map a column to marker size

sns.scatterplot(data=tips, x="total_bill", y="tip",
                hue="time", style="smoker", size="size")
plt.show()

One call, five variables on screen. Use this power sparingly — two semantics per chart is usually the readability limit.

Distributions: histplot

sns.histplot is matplotlib's histogram plus statistical extras — most usefully, an optional smoothed density curve (KDE) and per-group overlays with hue:

sns.histplot(data=tips, x="total_bill", bins=25, kde=True)
plt.show()

# overlay distributions per group
sns.histplot(data=tips, x="total_bill", hue="time", bins=25)
plt.show()

Categories: boxplot, violinplot, countplot

For a numeric variable split by category, seaborn's category plots are the biggest time-savers.

A boxplot summarizes each group with its median, quartile box, whiskers, and outlier dots:

sns.boxplot(data=tips, x="day", y="total_bill")
plt.show()

A violin plot replaces the box with the full density shape — better when groups might be bimodal (two bumps), which a box would hide:

sns.violinplot(data=tips, x="day", y="total_bill", hue="sex", split=True)
plt.show()

A countplot is a bar chart of frequencies — value_counts() as a picture:

sns.countplot(data=tips, x="day", hue="smoker")
plt.show()

Correlation heatmaps

A heatmap paints a matrix as colored cells. Its killer use case: the correlation matrix of your numeric columns, which shows every pairwise relationship at once:

penguins = sns.load_dataset("penguins")
 
corr = penguins.select_dtypes("number").corr()
sns.heatmap(corr, annot=True, fmt=".2f", cmap="coolwarm", vmin=-1, vmax=1)
plt.show()

annot=True prints the numbers in each cell; vmin/vmax=(-1, 1) anchors the color scale so that white really means "no correlation".

Pairplot: the one-line dataset overview

sns.pairplot draws a grid of scatter plots for every pair of numeric columns (with distributions on the diagonal). It's the classic first move in exploratory data analysis:

sns.pairplot(penguins, hue="species", vars=["bill_length_mm", "flipper_length_mm", "body_mass_g"])
plt.show()

On the penguins dataset this instantly reveals that the three species form separable clusters — exactly the kind of insight you want before any modeling.

Themes

One call restyles every subsequent chart — including plain matplotlib ones, since seaborn just configures matplotlib underneath:

sns.set_theme(style="whitegrid", palette="deep")   # do this once, at the top
sns.boxplot(data=tips, x="day", y="total_bill")
plt.show()

Other styles worth trying: "darkgrid" (the default), "white", "ticks".

In-browser practice (plain matplotlib)

You can approximate seaborn's two signature moves with matplotlib. First, a hue-style scatter — loop over groups, one color per group:

And a boxplot comparing groups — matplotlib's boxplot takes a list of arrays, one per category:

Same insight as sns.boxplot — weekend bills run higher and spread wider — just with a little more code. That's the trade in a nutshell.

That wraps up the data toolkit — you can now load, clean, query, and visualize data, which is everything you need to start the machine learning course.