Data Visualization with Matplotlib

Matplotlib is the plot you have seen a thousand times without noticing — loss curves in ML papers, figures in journal articles, the quick sanity check in someone's notebook. It's not the prettiest library, but it's the universal language of scientific plots — and learning the fig, ax = plt.subplots() pattern unlocks Seaborn, Plotly, and every other library built on top.

Learning Objectives

After this lesson, you will be able to:

Create line plots, scatter plots, bar charts, and histograms with Matplotlib
Customize plots with titles, labels, colors, legends, and annotations
Build multi-panel figures using subplots for side-by-side comparisons
Know when to use each chart type for different kinds of data
Understand how Seaborn extends Matplotlib for statistical visualizations

Prerequisites

Before You Start

Review

#Your First Plot

Matplotlib is Python's most widely used plotting library. The pyplot module provides a MATLAB-like interface:

pythonrunnable cell

import matplotlib.pyplot as plt
import numpy as np

# Generate data
x = np.linspace(0, 10, 100)  # 100 points from 0 to 10
y = np.sin(x)

# Create a line plot
plt.figure(figsize=(8, 4))  # width x height in inches
plt.plot(x, y)
plt.title("A Simple Sine Wave")
plt.xlabel("x")
plt.ylabel("sin(x)")
plt.grid(True)
plt.show()

#Adding Style and Detail

pythonreference · read-only

x = np.linspace(0, 2 * np.pi, 100)

plt.figure(figsize=(10, 5))
plt.plot(x, np.sin(x), label="sin(x)", color="blue", linewidth=2)
plt.plot(x, np.cos(x), label="cos(x)", color="red", linewidth=2, linestyle="--")
plt.plot(x, np.sin(x) + np.cos(x), label="sin(x) + cos(x)", color="purple", alpha=0.7)

plt.title("Trigonometric Functions", fontsize=16, fontweight="bold")
plt.xlabel("x (radians)", fontsize=12)
plt.ylabel("y", fontsize=12)
plt.legend(fontsize=11)
plt.grid(True, alpha=0.3)
plt.axhline(y=0, color="black", linewidth=0.5)  # horizontal line at y=0

plt.tight_layout()
plt.show()

Key parameters:

label — text for the legend
color — line color (names, hex codes, or RGB tuples)
linewidth — line thickness
linestyle — "-" solid, "--" dashed, ":" dotted, "-." dash-dot
alpha — transparency (0.0 = invisible, 1.0 = opaque)

#Scatter Plots

Scatter plots show the relationship between two numerical variables:

pythonrunnable cell

import numpy as np
import matplotlib.pyplot as plt

# Generate correlated data
np.random.seed(42)
hours_studied = np.random.uniform(1, 10, 50)
scores = 40 + 5 * hours_studied + np.random.normal(0, 5, 50)
scores = np.clip(scores, 0, 100)

plt.figure(figsize=(8, 6))
plt.scatter(hours_studied, scores, c="steelblue", alpha=0.7, edgecolors="white", s=80)

# Add a trend line
z = np.polyfit(hours_studied, scores, 1)
p = np.poly1d(z)
x_line = np.linspace(1, 10, 100)
plt.plot(x_line, p(x_line), "r--", linewidth=2, label=f"Trend: y = {z[0]:.1f}x + {z[1]:.1f}")

plt.title("Study Hours vs Test Scores", fontsize=14)
plt.xlabel("Hours Studied", fontsize=12)
plt.ylabel("Test Score", fontsize=12)
plt.legend()
plt.grid(True, alpha=0.3)
plt.tight_layout()
plt.show()

#Color-Coded Scatter Plots

pythonreference · read-only

# Color by a third variable
np.random.seed(42)
n = 100
math_scores = np.random.normal(75, 12, n)
science_scores = np.random.normal(78, 10, n)
attendance = np.random.normal(85, 10, n).clip(50, 100)

plt.figure(figsize=(8, 6))
scatter = plt.scatter(
    math_scores, science_scores,
    c=attendance,        # color by attendance
    cmap="RdYlGn",       # red-yellow-green colormap
    s=60, alpha=0.8, edgecolors="white"
)
plt.colorbar(scatter, label="Attendance %")
plt.title("Math vs Science Scores (colored by Attendance)")
plt.xlabel("Math Score")
plt.ylabel("Science Score")
plt.grid(True, alpha=0.3)
plt.tight_layout()
plt.show()

#Bar Charts and Histograms

#Bar Charts — Comparing Categories

pythonreference · read-only

subjects = ["Math", "Science", "English", "History", "Art"]
averages = [78.5, 82.3, 80.1, 75.8, 88.2]
colors = ["#3b82f6", "#22c55e", "#f59e0b", "#ef4444", "#a855f7"]

plt.figure(figsize=(8, 5))
bars = plt.bar(subjects, averages, color=colors, edgecolor="white", width=0.6)

# Add value labels on top of each bar
for bar, val in zip(bars, averages):
    plt.text(bar.get_x() + bar.get_width() / 2, bar.get_height() + 0.5,
             f"{val}", ha="center", fontsize=11, fontweight="bold")

plt.title("Average Scores by Subject", fontsize=14)
plt.ylabel("Average Score", fontsize=12)
plt.ylim(0, 100)
plt.grid(axis="y", alpha=0.3)
plt.tight_layout()
plt.show()

#Grouped Bar Charts

pythonreference · read-only

classes = ["9A", "9B", "9C"]
math_avgs = [82, 76, 79]
sci_avgs = [85, 80, 83]
eng_avgs = [78, 81, 77]

x = np.arange(len(classes))
width = 0.25

plt.figure(figsize=(8, 5))
plt.bar(x - width, math_avgs, width, label="Math", color="#3b82f6")
plt.bar(x, sci_avgs, width, label="Science", color="#22c55e")
plt.bar(x + width, eng_avgs, width, label="English", color="#f59e0b")

plt.title("Subject Averages by Class", fontsize=14)
plt.xlabel("Class")
plt.ylabel("Average Score")
plt.xticks(x, classes)
plt.legend()
plt.ylim(0, 100)
plt.grid(axis="y", alpha=0.3)
plt.tight_layout()
plt.show()

#Histograms — Seeing Distributions

pythonreference · read-only

np.random.seed(42)
scores = np.random.normal(75, 12, 500)  # 500 students

plt.figure(figsize=(8, 5))
plt.hist(scores, bins=25, color="steelblue", edgecolor="white", alpha=0.8)

# Add mean and median lines
mean_score = np.mean(scores)
median_score = np.median(scores)
plt.axvline(mean_score, color="red", linestyle="--", linewidth=2, label=f"Mean: {mean_score:.1f}")
plt.axvline(median_score, color="orange", linestyle="-.", linewidth=2, label=f"Median: {median_score:.1f}")

plt.title("Distribution of Student Scores (n=500)", fontsize=14)
plt.xlabel("Score", fontsize=12)
plt.ylabel("Frequency", fontsize=12)
plt.legend(fontsize=11)
plt.grid(axis="y", alpha=0.3)
plt.tight_layout()
plt.show()

What Do You Think?

When should you use a histogram instead of a bar chart?

Intuition

Build Your Intuition

Think of Figure as the canvas and Axes as the individual painting on that canvas.

Matplotlib has two styles. The pyplot (stateful) style tracks a "current figure" globally — convenient for single plots but fragile when you have multiple plots or functions that share state:

pythonreference · read-only

# pyplot style — the "implicit" API
plt.plot(x, y)          # always targets the "current" axes
plt.title("My Plot")    # same implicit target

The object-oriented (OO) style is explicit — you hold references to the Figure and each Axes object:

pythonreference · read-only

# OO style — the "explicit" API (always prefer this)
fig, ax = plt.subplots()    # fig = canvas, ax = one plot on it
ax.plot(x, y)               # unambiguous: we are plotting on THIS ax
ax.set_title("My Plot")     # same ax — no global state involved
fig.savefig("out.png")      # save THIS figure

With subplots you get multiple Axes objects: fig, axes = plt.subplots(2, 2) gives you a 2×2 array of Axes. Each ax is independent — setting the title on axes[0, 0] does not affect axes[1, 1]. This is why every professional Matplotlib code you will read in the wild uses fig, ax = plt.subplots() rather than plt.plot().

#Subplots — Multi-Panel Figures

Real analysis often requires multiple plots side by side:

pythonrunnable cell

import numpy as np
import matplotlib.pyplot as plt

np.random.seed(42)
n = 200
math = np.random.normal(75, 12, n).clip(0, 100)
science = np.random.normal(78, 10, n).clip(0, 100)
english = np.random.normal(80, 11, n).clip(0, 100)

fig, axes = plt.subplots(2, 2, figsize=(12, 10))
fig.suptitle("Student Performance Dashboard", fontsize=16, fontweight="bold")

# Plot 1: Score distributions (overlapping histograms)
ax1 = axes[0, 0]
ax1.hist(math, bins=20, alpha=0.6, label="Math", color="#3b82f6")
ax1.hist(science, bins=20, alpha=0.6, label="Science", color="#22c55e")
ax1.hist(english, bins=20, alpha=0.6, label="English", color="#f59e0b")
ax1.set_title("Score Distributions")
ax1.set_xlabel("Score")
ax1.set_ylabel("Frequency")
ax1.legend()
ax1.grid(axis="y", alpha=0.3)

# Plot 2: Math vs Science scatter
ax2 = axes[0, 1]
ax2.scatter(math, science, alpha=0.5, c="steelblue", edgecolors="white", s=40)
ax2.set_title("Math vs Science Scores")
ax2.set_xlabel("Math Score")
ax2.set_ylabel("Science Score")
ax2.grid(True, alpha=0.3)

# Plot 3: Average by subject (bar chart)
ax3 = axes[1, 0]
subjects = ["Math", "Science", "English"]
means = [np.mean(math), np.mean(science), np.mean(english)]
stds = [np.std(math), np.std(science), np.std(english)]
bars = ax3.bar(subjects, means, color=["#3b82f6", "#22c55e", "#f59e0b"],
               edgecolor="white", yerr=stds, capsize=5)
ax3.set_title("Average Scores (with Std Dev)")
ax3.set_ylabel("Score")
ax3.set_ylim(0, 100)
ax3.grid(axis="y", alpha=0.3)

# Plot 4: Box plots for comparison
ax4 = axes[1, 1]
bp = ax4.boxplot([math, science, english], tick_labels=subjects,
                  patch_artist=True, medianprops={"color": "black", "linewidth": 2})
# (Matplotlib 3.9+: use `tick_labels=`. The older `labels=` is deprecated.)
colors = ["#3b82f6", "#22c55e", "#f59e0b"]
for patch, color in zip(bp["boxes"], colors):
    patch.set_facecolor(color)
    patch.set_alpha(0.7)
ax4.set_title("Score Distribution (Box Plots)")
ax4.set_ylabel("Score")
ax4.grid(axis="y", alpha=0.3)

plt.tight_layout()
plt.show()

#Subplot Syntax

pythonreference · read-only

# Method 1: fig, axes = plt.subplots(rows, cols)
fig, axes = plt.subplots(2, 3, figsize=(15, 8))  # 2 rows, 3 columns
axes[0, 0].plot(x, y)    # access by row, col index
axes[1, 2].bar(...)       # bottom-right plot

# Method 2: Single row or column
fig, (ax1, ax2, ax3) = plt.subplots(1, 3, figsize=(15, 4))
ax1.plot(x, y)
ax2.scatter(x, y)
ax3.hist(y)

# Always call tight_layout() to prevent overlapping
plt.tight_layout()

#Customization and Best Practices

#Choosing the Right Chart

Chart Type	Best For	Example
Line plot	Trends over time or continuous sequences	Training loss over epochs
Scatter plot	Relationship between two numerical variables	Height vs weight
Bar chart	Comparing categories	Average score per subject
Histogram	Distribution of a single numerical variable	Score distribution
Box plot	Comparing distributions across groups	Scores by class section
Pie chart	Part-to-whole relationships (use sparingly)	Market share
Heatmap	Correlation matrices or 2D data	Feature correlations

Build Your Own ChartInteractive

Loading visualization...

#Saving Plots

pythonreference · read-only

# Save to file (always save BEFORE plt.show())
fig, ax = plt.subplots(figsize=(8, 5))
ax.plot([1, 2, 3], [4, 5, 6])
ax.set_title("My Plot")

fig.savefig("my_plot.png", dpi=150, bbox_inches="tight")   # PNG
fig.savefig("my_plot.pdf", bbox_inches="tight")             # PDF (vector)
fig.savefig("my_plot.svg", bbox_inches="tight")             # SVG (vector)
plt.show()

#Matplotlib Style Sheets

pythonreference · read-only

# Use a built-in style
plt.style.use("seaborn-v0_8-whitegrid")  # clean, modern look

# Other popular styles:
# plt.style.use("ggplot")         # R's ggplot2 look
# plt.style.use("dark_background") # dark theme
# plt.style.use("fivethirtyeight") # FiveThirtyEight blog style

#A Taste of Seaborn

Seaborn is built on top of Matplotlib and provides beautiful statistical plots with minimal code:

pythonreference · read-only

import seaborn as sns
import pandas as pd

# Seaborn works beautifully with Pandas DataFrames
df = pd.DataFrame({
    "student": [f"S{i}" for i in range(100)],
    "hours_studied": np.random.uniform(1, 10, 100),
    "score": np.random.normal(75, 12, 100).clip(0, 100).round(),
    "class": np.random.choice(["9A", "9B", "9C"], 100),
})

# Scatter with regression line -- one line of code!
sns.lmplot(data=df, x="hours_studied", y="score", hue="class",
           height=5, aspect=1.3)
plt.title("Study Hours vs Score by Class")
plt.show()

# Distribution plot
fig, axes = plt.subplots(1, 2, figsize=(12, 4))
sns.histplot(data=df, x="score", hue="class", kde=True, ax=axes[0])
axes[0].set_title("Score Distribution by Class")

sns.boxplot(data=df, x="class", y="score", hue="class", palette="Set2", legend=False, ax=axes[1])
axes[1].set_title("Score Box Plots by Class")
plt.tight_layout()
plt.show()

# Heatmap -- great for correlation matrices
numeric_cols = df[["hours_studied", "score"]].copy()
numeric_cols["attendance"] = np.random.normal(85, 10, 100)
corr = numeric_cols.corr()

plt.figure(figsize=(6, 4))
sns.heatmap(corr, annot=True, cmap="RdBu_r", center=0, fmt=".2f",
            square=True, linewidths=1)
plt.title("Feature Correlation Matrix")
plt.tight_layout()
plt.show()

Seaborn is not a replacement for Matplotlib — it is built on top of it. You often use both together: Seaborn for high-level statistical plots, Matplotlib for customization.

#Code Playground

pythonplayground.py · Pyodide

# Data Visualization Concepts in Python
# Data preparation and analysis that feeds into matplotlib plots
import math
import random

# === DATASET: Monthly Sales Data ===
months = ["Jan", "Feb", "Mar", "Apr", "May", "Jun",
        "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"]
sales_2024 = [120, 135, 150, 145, 160, 175, 190, 185, 170, 195, 210, 230]
sales_2025 = [130, 140, 155, 165, 175, 180, 200, 195, 185, 205, 225, 250]

# === LINE PLOT DATA (trends over time) ===
print("=== LINE PLOT: Monthly Sales Trend ===")
print(f"{'Month':<6}| {'2024':<6}| {'2025':<6}| Change")
print("-" * 40)
for i, m in enumerate(months):
  change = sales_2025[i] - sales_2024[i]
  sign = "+" if change > 0 else ""
  print(f"{m:<6}| {sales_2024[i]:<6}| {sales_2025[i]:<6}| {sign}{change}")

# === BAR CHART DATA (category comparison) ===
print("\n=== BAR CHART: Quarterly Totals ===")
def quarterly_total(monthly):
  return [sum(monthly[i:i+3]) for i in range(0, 12, 3)]

q2024 = quarterly_total(sales_2024)
q2025 = quarterly_total(sales_2025)
for q, q24, q25 in zip(["Q1", "Q2", "Q3", "Q4"], q2024, q2025):
  print(f"{q} 2024: {'#' * round(q24 / 20)} {q24}")
  print(f"{q} 2025: {'#' * round(q25 / 20)} {q25}")

# === HISTOGRAM DATA (distribution) ===
print("\n=== HISTOGRAM: Score Distribution ===")
random.seed(42)
scores = [max(0, min(100, round(random.gauss(75, 12)))) for _ in range(200)]

bins: dict[str, int] = {}
for s in scores:
  b = (s // 10) * 10
  label = f"{b}-{b+9}"
  bins[label] = bins.get(label, 0) + 1

print(f"{'Range':<9}| {'Count':<6}| Bar")
for rng_label, count in sorted(bins.items()):
  print(f"{rng_label:<9}| {count:<6}| {'*' * count}")

# === SCATTER PLOT DATA (correlation) ===
print("\n=== SCATTER: Hours Studied vs Score ===")
students = []
for _ in range(20):
  hours = round(1 + random.random() * 9, 1)
  score = max(0, min(100, round(40 + 5 * hours + (random.random() - 0.5) * 20)))
  students.append({"hours": hours, "score": score})

# Correlation using numpy-free math
h_vals = [s["hours"] for s in students]
s_vals = [s["score"] for s in students]
h_mean = sum(h_vals) / len(h_vals)
s_mean = sum(s_vals) / len(s_vals)
num = sum((h - h_mean) * (s - s_mean) for h, s in zip(h_vals, s_vals))
d1 = sum((h - h_mean) ** 2 for h in h_vals)
d2 = sum((s - s_mean) ** 2 for s in s_vals)
corr = num / math.sqrt(d1 * d2)

for st in students:
  print(f"  Hours: {st['hours']:<5} Score: {st['score']}")
print(f"\nCorrelation (r): {corr:.3f}")
print("Strong positive correlation!" if corr > 0.5 else "Weak correlation")

# CHALLENGES:
# 1. Calculate year-over-year growth percentage for each month
# 2. Find the month with the highest growth rate
# 3. Create a "text sparkline" showing the trend (use chars like _ - ^ for low/mid/high)
# 4. Calculate the moving average of sales_2025 with a window of 3 months

Tests · Prepare data for visualization, calculate statistics, and explore trends!

#ML-Specific Plots You'll Use Every Day

pythonrunnable cell

import matplotlib.pyplot as plt
import numpy as np

# 1. Training loss curve — the most important ML plot
epochs = range(1, 21)
train_loss = [1.0 * (0.85 ** e) + np.random.normal(0, 0.02) for e in epochs]
val_loss   = [1.0 * (0.87 ** e) + np.random.normal(0, 0.03) for e in epochs]

plt.figure(figsize=(8, 4))
plt.plot(epochs, train_loss, 'b-o', label='Training Loss', markersize=4)
plt.plot(epochs, val_loss,   'r-o', label='Validation Loss', markersize=4)
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.title('Training Progress — watch for validation loss rising (overfitting!)')
plt.legend()
plt.grid(alpha=0.3)
plt.show()

# 2. Feature importance bar chart
features   = ['Credit Score', 'Income', 'Loan Amount', 'Employment Yrs']
importance = [0.38, 0.27, 0.22, 0.13]

colors = ['#818cf8', '#22d3ee', '#34d399', '#fbbf24']
plt.barh(features, importance, color=colors)
plt.xlabel('Importance Score')
plt.title('Which features does the model rely on?')
plt.tight_layout()
plt.show()

Reading the loss curve: If training loss keeps dropping but validation loss starts rising — that's overfitting. The model is memorising training data instead of learning the pattern. This single chart drives most hyperparameter decisions.

#Key Takeaways

plt.plot() for lines, plt.scatter() for relationships, plt.bar() for categories, plt.hist() for distributions — choosing the right chart type is the most important visualization decision
Always label your plots — every chart needs a title, axis labels, and a legend (if multiple series). Unlabeled plots are useless to everyone, including your future self
Subplots let you build dashboards — fig, axes = plt.subplots(rows, cols) creates a grid of plots. Use this to show multiple perspectives of the same data side by side
Seaborn makes statistical plots effortless — for scatter with regression (lmplot), distributions (histplot), correlations (heatmap), and box plots (boxplot), Seaborn produces beautiful results in one line
Save before you show — fig.savefig("plot.png", dpi=150) must come before plt.show(). After show(), the figure object is cleared and saving produces a blank image

Quick Check1 / 5

Which chart type is best for showing the distribution of a single numerical variable?

Data Visualization with Matplotlib

Learning Objectives

After this lesson, you will be able to:

Create line plots, scatter plots, bar charts, and histograms with Matplotlib
Customize plots with titles, labels, colors, legends, and annotations
Build multi-panel figures using subplots for side-by-side comparisons
Know when to use each chart type for different kinds of data
Understand how Seaborn extends Matplotlib for statistical visualizations

Prerequisites

Before You Start

Review

#Your First Plot

Matplotlib is Python's most widely used plotting library. The pyplot module provides a MATLAB-like interface:

pythonrunnable cell

import matplotlib.pyplot as plt
import numpy as np

# Generate data
x = np.linspace(0, 10, 100)  # 100 points from 0 to 10
y = np.sin(x)

# Create a line plot
plt.figure(figsize=(8, 4))  # width x height in inches
plt.plot(x, y)
plt.title("A Simple Sine Wave")
plt.xlabel("x")
plt.ylabel("sin(x)")
plt.grid(True)
plt.show()

#Adding Style and Detail

pythonreference · read-only

x = np.linspace(0, 2 * np.pi, 100)

plt.figure(figsize=(10, 5))
plt.plot(x, np.sin(x), label="sin(x)", color="blue", linewidth=2)
plt.plot(x, np.cos(x), label="cos(x)", color="red", linewidth=2, linestyle="--")
plt.plot(x, np.sin(x) + np.cos(x), label="sin(x) + cos(x)", color="purple", alpha=0.7)

plt.title("Trigonometric Functions", fontsize=16, fontweight="bold")
plt.xlabel("x (radians)", fontsize=12)
plt.ylabel("y", fontsize=12)
plt.legend(fontsize=11)
plt.grid(True, alpha=0.3)
plt.axhline(y=0, color="black", linewidth=0.5)  # horizontal line at y=0

plt.tight_layout()
plt.show()

Key parameters:

label — text for the legend
color — line color (names, hex codes, or RGB tuples)
linewidth — line thickness
linestyle — "-" solid, "--" dashed, ":" dotted, "-." dash-dot
alpha — transparency (0.0 = invisible, 1.0 = opaque)

#Scatter Plots

Scatter plots show the relationship between two numerical variables:

pythonrunnable cell

import numpy as np
import matplotlib.pyplot as plt

# Generate correlated data
np.random.seed(42)
hours_studied = np.random.uniform(1, 10, 50)
scores = 40 + 5 * hours_studied + np.random.normal(0, 5, 50)
scores = np.clip(scores, 0, 100)

plt.figure(figsize=(8, 6))
plt.scatter(hours_studied, scores, c="steelblue", alpha=0.7, edgecolors="white", s=80)

# Add a trend line
z = np.polyfit(hours_studied, scores, 1)
p = np.poly1d(z)
x_line = np.linspace(1, 10, 100)
plt.plot(x_line, p(x_line), "r--", linewidth=2, label=f"Trend: y = {z[0]:.1f}x + {z[1]:.1f}")

plt.title("Study Hours vs Test Scores", fontsize=14)
plt.xlabel("Hours Studied", fontsize=12)
plt.ylabel("Test Score", fontsize=12)
plt.legend()
plt.grid(True, alpha=0.3)
plt.tight_layout()
plt.show()

#Color-Coded Scatter Plots

pythonreference · read-only

# Color by a third variable
np.random.seed(42)
n = 100
math_scores = np.random.normal(75, 12, n)
science_scores = np.random.normal(78, 10, n)
attendance = np.random.normal(85, 10, n).clip(50, 100)

plt.figure(figsize=(8, 6))
scatter = plt.scatter(
    math_scores, science_scores,
    c=attendance,        # color by attendance
    cmap="RdYlGn",       # red-yellow-green colormap
    s=60, alpha=0.8, edgecolors="white"
)
plt.colorbar(scatter, label="Attendance %")
plt.title("Math vs Science Scores (colored by Attendance)")
plt.xlabel("Math Score")
plt.ylabel("Science Score")
plt.grid(True, alpha=0.3)
plt.tight_layout()
plt.show()

#Bar Charts and Histograms

#Bar Charts — Comparing Categories

pythonreference · read-only

subjects = ["Math", "Science", "English", "History", "Art"]
averages = [78.5, 82.3, 80.1, 75.8, 88.2]
colors = ["#3b82f6", "#22c55e", "#f59e0b", "#ef4444", "#a855f7"]

plt.figure(figsize=(8, 5))
bars = plt.bar(subjects, averages, color=colors, edgecolor="white", width=0.6)

# Add value labels on top of each bar
for bar, val in zip(bars, averages):
    plt.text(bar.get_x() + bar.get_width() / 2, bar.get_height() + 0.5,
             f"{val}", ha="center", fontsize=11, fontweight="bold")

plt.title("Average Scores by Subject", fontsize=14)
plt.ylabel("Average Score", fontsize=12)
plt.ylim(0, 100)
plt.grid(axis="y", alpha=0.3)
plt.tight_layout()
plt.show()

#Grouped Bar Charts

pythonreference · read-only

classes = ["9A", "9B", "9C"]
math_avgs = [82, 76, 79]
sci_avgs = [85, 80, 83]
eng_avgs = [78, 81, 77]

x = np.arange(len(classes))
width = 0.25

plt.figure(figsize=(8, 5))
plt.bar(x - width, math_avgs, width, label="Math", color="#3b82f6")
plt.bar(x, sci_avgs, width, label="Science", color="#22c55e")
plt.bar(x + width, eng_avgs, width, label="English", color="#f59e0b")

plt.title("Subject Averages by Class", fontsize=14)
plt.xlabel("Class")
plt.ylabel("Average Score")
plt.xticks(x, classes)
plt.legend()
plt.ylim(0, 100)
plt.grid(axis="y", alpha=0.3)
plt.tight_layout()
plt.show()

#Histograms — Seeing Distributions

pythonreference · read-only

np.random.seed(42)
scores = np.random.normal(75, 12, 500)  # 500 students

plt.figure(figsize=(8, 5))
plt.hist(scores, bins=25, color="steelblue", edgecolor="white", alpha=0.8)

# Add mean and median lines
mean_score = np.mean(scores)
median_score = np.median(scores)
plt.axvline(mean_score, color="red", linestyle="--", linewidth=2, label=f"Mean: {mean_score:.1f}")
plt.axvline(median_score, color="orange", linestyle="-.", linewidth=2, label=f"Median: {median_score:.1f}")

plt.title("Distribution of Student Scores (n=500)", fontsize=14)
plt.xlabel("Score", fontsize=12)
plt.ylabel("Frequency", fontsize=12)
plt.legend(fontsize=11)
plt.grid(axis="y", alpha=0.3)
plt.tight_layout()
plt.show()

What Do You Think?

When should you use a histogram instead of a bar chart?

Intuition

Build Your Intuition

Think of Figure as the canvas and Axes as the individual painting on that canvas.

Matplotlib has two styles. The pyplot (stateful) style tracks a "current figure" globally — convenient for single plots but fragile when you have multiple plots or functions that share state:

pythonreference · read-only

# pyplot style — the "implicit" API
plt.plot(x, y)          # always targets the "current" axes
plt.title("My Plot")    # same implicit target

The object-oriented (OO) style is explicit — you hold references to the Figure and each Axes object:

pythonreference · read-only

# OO style — the "explicit" API (always prefer this)
fig, ax = plt.subplots()    # fig = canvas, ax = one plot on it
ax.plot(x, y)               # unambiguous: we are plotting on THIS ax
ax.set_title("My Plot")     # same ax — no global state involved
fig.savefig("out.png")      # save THIS figure

#Subplots — Multi-Panel Figures

Real analysis often requires multiple plots side by side:

pythonrunnable cell

import numpy as np
import matplotlib.pyplot as plt

np.random.seed(42)
n = 200
math = np.random.normal(75, 12, n).clip(0, 100)
science = np.random.normal(78, 10, n).clip(0, 100)
english = np.random.normal(80, 11, n).clip(0, 100)

fig, axes = plt.subplots(2, 2, figsize=(12, 10))
fig.suptitle("Student Performance Dashboard", fontsize=16, fontweight="bold")

# Plot 1: Score distributions (overlapping histograms)
ax1 = axes[0, 0]
ax1.hist(math, bins=20, alpha=0.6, label="Math", color="#3b82f6")
ax1.hist(science, bins=20, alpha=0.6, label="Science", color="#22c55e")
ax1.hist(english, bins=20, alpha=0.6, label="English", color="#f59e0b")
ax1.set_title("Score Distributions")
ax1.set_xlabel("Score")
ax1.set_ylabel("Frequency")
ax1.legend()
ax1.grid(axis="y", alpha=0.3)

# Plot 2: Math vs Science scatter
ax2 = axes[0, 1]
ax2.scatter(math, science, alpha=0.5, c="steelblue", edgecolors="white", s=40)
ax2.set_title("Math vs Science Scores")
ax2.set_xlabel("Math Score")
ax2.set_ylabel("Science Score")
ax2.grid(True, alpha=0.3)

# Plot 3: Average by subject (bar chart)
ax3 = axes[1, 0]
subjects = ["Math", "Science", "English"]
means = [np.mean(math), np.mean(science), np.mean(english)]
stds = [np.std(math), np.std(science), np.std(english)]
bars = ax3.bar(subjects, means, color=["#3b82f6", "#22c55e", "#f59e0b"],
               edgecolor="white", yerr=stds, capsize=5)
ax3.set_title("Average Scores (with Std Dev)")
ax3.set_ylabel("Score")
ax3.set_ylim(0, 100)
ax3.grid(axis="y", alpha=0.3)

# Plot 4: Box plots for comparison
ax4 = axes[1, 1]
bp = ax4.boxplot([math, science, english], tick_labels=subjects,
                  patch_artist=True, medianprops={"color": "black", "linewidth": 2})
# (Matplotlib 3.9+: use `tick_labels=`. The older `labels=` is deprecated.)
colors = ["#3b82f6", "#22c55e", "#f59e0b"]
for patch, color in zip(bp["boxes"], colors):
    patch.set_facecolor(color)
    patch.set_alpha(0.7)
ax4.set_title("Score Distribution (Box Plots)")
ax4.set_ylabel("Score")
ax4.grid(axis="y", alpha=0.3)

plt.tight_layout()
plt.show()

#Subplot Syntax

pythonreference · read-only

# Method 1: fig, axes = plt.subplots(rows, cols)
fig, axes = plt.subplots(2, 3, figsize=(15, 8))  # 2 rows, 3 columns
axes[0, 0].plot(x, y)    # access by row, col index
axes[1, 2].bar(...)       # bottom-right plot

# Method 2: Single row or column
fig, (ax1, ax2, ax3) = plt.subplots(1, 3, figsize=(15, 4))
ax1.plot(x, y)
ax2.scatter(x, y)
ax3.hist(y)

# Always call tight_layout() to prevent overlapping
plt.tight_layout()

#Customization and Best Practices

#Choosing the Right Chart

Chart Type	Best For	Example
Line plot	Trends over time or continuous sequences	Training loss over epochs
Scatter plot	Relationship between two numerical variables	Height vs weight
Bar chart	Comparing categories	Average score per subject
Histogram	Distribution of a single numerical variable	Score distribution
Box plot	Comparing distributions across groups	Scores by class section
Pie chart	Part-to-whole relationships (use sparingly)	Market share
Heatmap	Correlation matrices or 2D data	Feature correlations

Build Your Own ChartInteractive

Loading visualization...

#Saving Plots

pythonreference · read-only

# Save to file (always save BEFORE plt.show())
fig, ax = plt.subplots(figsize=(8, 5))
ax.plot([1, 2, 3], [4, 5, 6])
ax.set_title("My Plot")

fig.savefig("my_plot.png", dpi=150, bbox_inches="tight")   # PNG
fig.savefig("my_plot.pdf", bbox_inches="tight")             # PDF (vector)
fig.savefig("my_plot.svg", bbox_inches="tight")             # SVG (vector)
plt.show()

#Matplotlib Style Sheets

pythonreference · read-only

# Use a built-in style
plt.style.use("seaborn-v0_8-whitegrid")  # clean, modern look

# Other popular styles:
# plt.style.use("ggplot")         # R's ggplot2 look
# plt.style.use("dark_background") # dark theme
# plt.style.use("fivethirtyeight") # FiveThirtyEight blog style

#A Taste of Seaborn

Seaborn is built on top of Matplotlib and provides beautiful statistical plots with minimal code:

pythonreference · read-only

import seaborn as sns
import pandas as pd

# Seaborn works beautifully with Pandas DataFrames
df = pd.DataFrame({
    "student": [f"S{i}" for i in range(100)],
    "hours_studied": np.random.uniform(1, 10, 100),
    "score": np.random.normal(75, 12, 100).clip(0, 100).round(),
    "class": np.random.choice(["9A", "9B", "9C"], 100),
})

# Scatter with regression line -- one line of code!
sns.lmplot(data=df, x="hours_studied", y="score", hue="class",
           height=5, aspect=1.3)
plt.title("Study Hours vs Score by Class")
plt.show()

# Distribution plot
fig, axes = plt.subplots(1, 2, figsize=(12, 4))
sns.histplot(data=df, x="score", hue="class", kde=True, ax=axes[0])
axes[0].set_title("Score Distribution by Class")

sns.boxplot(data=df, x="class", y="score", hue="class", palette="Set2", legend=False, ax=axes[1])
axes[1].set_title("Score Box Plots by Class")
plt.tight_layout()
plt.show()

# Heatmap -- great for correlation matrices
numeric_cols = df[["hours_studied", "score"]].copy()
numeric_cols["attendance"] = np.random.normal(85, 10, 100)
corr = numeric_cols.corr()

plt.figure(figsize=(6, 4))
sns.heatmap(corr, annot=True, cmap="RdBu_r", center=0, fmt=".2f",
            square=True, linewidths=1)
plt.title("Feature Correlation Matrix")
plt.tight_layout()
plt.show()

Seaborn is not a replacement for Matplotlib — it is built on top of it. You often use both together: Seaborn for high-level statistical plots, Matplotlib for customization.

#Code Playground

pythonplayground.py · Pyodide

# Data Visualization Concepts in Python
# Data preparation and analysis that feeds into matplotlib plots
import math
import random

# === BAR CHART DATA (category comparison) ===
print("\n=== BAR CHART: Quarterly Totals ===")
def quarterly_total(monthly):
  return [sum(monthly[i:i+3]) for i in range(0, 12, 3)]

# === HISTOGRAM DATA (distribution) ===
print("\n=== HISTOGRAM: Score Distribution ===")
random.seed(42)
scores = [max(0, min(100, round(random.gauss(75, 12)))) for _ in range(200)]

bins: dict[str, int] = {}
for s in scores:
  b = (s // 10) * 10
  label = f"{b}-{b+9}"
  bins[label] = bins.get(label, 0) + 1

print(f"{'Range':<9}| {'Count':<6}| Bar")
for rng_label, count in sorted(bins.items()):
  print(f"{rng_label:<9}| {count:<6}| {'*' * count}")

for st in students:
  print(f"  Hours: {st['hours']:<5} Score: {st['score']}")
print(f"\nCorrelation (r): {corr:.3f}")
print("Strong positive correlation!" if corr > 0.5 else "Weak correlation")

Tests · Prepare data for visualization, calculate statistics, and explore trends!

#ML-Specific Plots You'll Use Every Day

pythonrunnable cell

import matplotlib.pyplot as plt
import numpy as np

# 1. Training loss curve — the most important ML plot
epochs = range(1, 21)
train_loss = [1.0 * (0.85 ** e) + np.random.normal(0, 0.02) for e in epochs]
val_loss   = [1.0 * (0.87 ** e) + np.random.normal(0, 0.03) for e in epochs]

plt.figure(figsize=(8, 4))
plt.plot(epochs, train_loss, 'b-o', label='Training Loss', markersize=4)
plt.plot(epochs, val_loss,   'r-o', label='Validation Loss', markersize=4)
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.title('Training Progress — watch for validation loss rising (overfitting!)')
plt.legend()
plt.grid(alpha=0.3)
plt.show()

# 2. Feature importance bar chart
features   = ['Credit Score', 'Income', 'Loan Amount', 'Employment Yrs']
importance = [0.38, 0.27, 0.22, 0.13]

colors = ['#818cf8', '#22d3ee', '#34d399', '#fbbf24']
plt.barh(features, importance, color=colors)
plt.xlabel('Importance Score')
plt.title('Which features does the model rely on?')
plt.tight_layout()
plt.show()

#Key Takeaways

plt.plot() for lines, plt.scatter() for relationships, plt.bar() for categories, plt.hist() for distributions — choosing the right chart type is the most important visualization decision
Always label your plots — every chart needs a title, axis labels, and a legend (if multiple series). Unlabeled plots are useless to everyone, including your future self
Subplots let you build dashboards — fig, axes = plt.subplots(rows, cols) creates a grid of plots. Use this to show multiple perspectives of the same data side by side
Seaborn makes statistical plots effortless — for scatter with regression (lmplot), distributions (histplot), correlations (heatmap), and box plots (boxplot), Seaborn produces beautiful results in one line
Save before you show — fig.savefig("plot.png", dpi=150) must come before plt.show(). After show(), the figure object is cleared and saving produces a blank image

Quick Check1 / 5

Which chart type is best for showing the distribution of a single numerical variable?

Data Visualization with Matplotlib

#Your First Plot

#Adding Style and Detail

#Scatter Plots

#Color-Coded Scatter Plots

#Bar Charts and Histograms

#Bar Charts — Comparing Categories

#Grouped Bar Charts

#Histograms — Seeing Distributions

#Subplots — Multi-Panel Figures

#Subplot Syntax

#Customization and Best Practices

#Choosing the Right Chart

#Saving Plots

#Matplotlib Style Sheets

#A Taste of Seaborn

#Code Playground

The Two Matplotlib APIs — pyplot vs Object-Oriented

#ML-Specific Plots You'll Use Every Day

#Key Takeaways

Your Reflection

Continue Learning

Discussion

Data Visualization with Matplotlib

#Your First Plot

#Adding Style and Detail

#Scatter Plots

#Color-Coded Scatter Plots

#Bar Charts and Histograms

#Bar Charts — Comparing Categories

#Grouped Bar Charts

#Histograms — Seeing Distributions

#Subplots — Multi-Panel Figures

#Subplot Syntax

#Customization and Best Practices

#Choosing the Right Chart

#Saving Plots

#Matplotlib Style Sheets

#A Taste of Seaborn

#Code Playground

The Two Matplotlib APIs — pyplot vs Object-Oriented

#ML-Specific Plots You'll Use Every Day

#Key Takeaways

Your Reflection

Continue Learning

Discussion