Data Visualization

A chart can communicate in one glance what a spreadsheet takes paragraphs to say. This chapter covers Matplotlib for precise, customizable plots and Seaborn for attractive statistical charts. The goal is not just to produce pictures, but to produce pictures that honestly and clearly tell the story of the data.

Matplotlib Basics

Matplotlib is the oldest and most flexible plotting library in Python. Its core idea is a figure (the whole canvas) containing one or more axes (the individual plots drawn on it).

Figure and Axes

import matplotlib.pyplot as plt

fig, ax = plt.subplots(figsize=(8, 5))
ax.plot([1, 2, 3, 4], [2, 4, 1, 5])
ax.set_title("A Simple Line")
ax.set_xlabel("Index")
ax.set_ylabel("Value")
plt.show()

Line, Bar, Scatter, and Histogram Plots

import numpy as np

x = np.linspace(0, 10, 100)
fig, axes = plt.subplots(2, 2, figsize=(10, 8))

axes[0, 0].plot(x, np.sin(x), color="purple")        # line
axes[0, 0].set_title("Line")

axes[0, 1].bar(["A", "B", "C"], [3, 7, 5])           # bar
axes[0, 1].set_title("Bar")

rng = np.random.default_rng(42)
axes[1, 0].scatter(rng.normal(size=50), rng.normal(size=50))  # scatter
axes[1, 0].set_title("Scatter")

axes[1, 1].hist(rng.normal(size=500), bins=20)       # histogram
axes[1, 1].set_title("Histogram")

plt.tight_layout()
plt.show()

Labels, Legends, and Saving Figures

fig, ax = plt.subplots()
ax.plot(x, np.sin(x), label="sin")
ax.plot(x, np.cos(x), label="cos")
ax.legend()
ax.set_xlabel("Angle (radians)")
ax.set_ylabel("Value")
ax.set_title("Sine and Cosine")
fig.savefig("wave.png", dpi=150, bbox_inches="tight")

Seaborn: Statistical Plotting Made Easy

Seaborn works directly with Pandas DataFrames and adds themes and high-level statistical plots. It imports as sns and sits on top of Matplotlib, so every Matplotlib trick still works.

Style Themes

import seaborn as sns
import pandas as pd

sns.set_theme(style="whitegrid")
data = pd.DataFrame({
    "group": ["a"] * 50 + ["b"] * 50,
    "value": list(range(50)) + list(range(50, 100))
})

Distribution Plots: histplot

histplot (the modern replacement for the older distplot) draws a histogram and can overlay a kernel density estimate.

sns.histplot(data["value"], bins=20, kde=True)
plt.show()

Boxplots and Categorical Plots

sns.boxplot(data=data, x="group", y="value")
plt.title("Values by Group")
plt.show()

sns.catplot(data=data, x="group", y="value", kind="box")
sns.countplot(x="group", data=data)
plt.show()

Pairplots and Heatmaps

import numpy as np

wide = pd.DataFrame({
    "a": np.random.default_rng(1).normal(size=100),
    "b": np.random.default_rng(2).normal(size=100),
    "c": np.random.default_rng(3).normal(size=100),
})
sns.pairplot(wide)
plt.show()

corr = wide.corr()
sns.heatmap(corr, annot=True, cmap="coolwarm")
plt.title("Correlation Matrix")
plt.show()

Practical Tips for Readable Charts

• Use plt.tight_layout() so titles and labels never overlap between subplots.
• Choose accessible color palettes (Seaborn themes do this for you) and avoid red-green pairs.
• Limit histogram bins to a sensible count; too many bins produce noise, too few hide shape.
• Set your figure size before drawing with plt.subplots(figsize=(...)).
• Annotate the key insight with ax.annotate or a simple ax.text rather than making readers hunt for it.
• Always export at dpi=150 or higher so charts stay sharp when embedded.

Visualization Practice

Build a small end-to-end example that exercises everything above. We use the classic "tips" dataset that ships with Seaborn so the numbers are real.

tips = sns.load_dataset("tips")
print(tips.head())

fig, ax = plt.subplots(figsize=(8, 5))
sns.scatterplot(data=tips, x="total_bill", y="tip", hue="day", ax=ax)
ax.set_title("Tip vs Total Bill by Day")
fig.savefig("tips_scatter.png", dpi=150)
plt.show()

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