Matplotlib

Topic — The foundation library. Seaborn and pandas' own .plot() both sit on top of it, so its vocabulary is worth learning first.

import pandas as pd
import matplotlib.pyplot as plt

plt is the universal alias for matplotlib.pyplot. Every chart on this page is drawn from one DataFrame:

df = pd.DataFrame({
    "Student": ["Arun", "Bala", "Charan", "Divya", "Esha",
                "Farah", "Gokul", "Hari", "Indhu", "Jaya"],
    "Study_Hours": [2, 3, 5, 1, 4, 3, 4, 6, 2, 5],
    "Attendance": [70, 75, 90, 60, 85, 78, 88, 95, 72, 92],
    "Marks": [55, 62, 85, 45, 78, 68, 80, 95, 58, 88],
})

---

The pattern

Every chart follows the same four steps:

plt.figure()                          # 1. start a new figure
plt.plot(df["Student"], df["Marks"])  # 2. draw something
plt.title("Student Marks")            # 3. label it
plt.show()                            # 4. display it

Call	Does
plt.figure()	starts a fresh canvas
plt.plot() / .bar() / .hist() …	draws onto the current figure
plt.title() / .xlabel() / .ylabel()	labels
plt.show()	renders and clears

warning

plt.figure() matters more than it looks

Without it, every drawing call lands on the same figure, so your second chart is drawn on top of the first. plt.show() clears the current figure, so in a script where every block ends with show() you often get away with it — but the moment one block doesn't, charts merge.

Start every chart with plt.figure().

---

Line plot

plt.figure()
plt.plot(df["Student"], df["Marks"], marker="o")
plt.title("Student Marks")
plt.xlabel("Student")
plt.ylabel("Marks")
plt.show()

marker="o" puts a dot at each data point. Without it you get a bare line and can't tell where the actual observations are.

This chart is technically fine and analytically wrong

The x-axis is student names in dataset order. A line implies that consecutive points are connected — that Arun becomes Bala. There's no such progression.

A line plot needs an ordered x-axis: time, an index, a measured quantity. For comparing students, use a bar chart. See Choosing a Chart.

---

Bar chart

plt.figure()
plt.bar(df["Student"], df["Marks"])
plt.title("Marks of Students")
plt.xlabel("Student")
plt.ylabel("Marks")
plt.show()

This is the right chart for this data. Bars compare magnitudes across unordered categories, and the eye reads length accurately. Hari's 95 and Divya's 45 are immediately comparable.

Note the y-axis starts at zero — matplotlib's default for bar(), and the correct behaviour. A truncated axis makes small differences look enormous.

plt.barh() draws horizontal bars, which is better when labels are long.

---

Histogram

plt.figure()
plt.hist(df["Marks"], bins=5)
plt.title("Distribution of Marks")
plt.xlabel("Marks")
plt.ylabel("Number of Students")
plt.show()

A histogram takes one numeric column, divides its range into bins, and counts how many values fall in each. The y-axis is a count, not a value.

Compare with the bar chart above: same column, completely different question. The bar chart shows each student's mark; the histogram shows how marks are spread and loses track of who's who.

bins=5 splits the 45–95 range into five intervals of 10. Try bins=3 and bins=10 on the same data — the apparent shape changes, which is why the bin count is a choice you should make consciously rather than accept.

---

Scatter plot

plt.figure()
plt.scatter(df["Study_Hours"], df["Marks"])
plt.title("Study Hours vs Marks")
plt.xlabel("Study Hours")
plt.ylabel("Marks")
plt.show()

One point per student, study hours on x, marks on y. The upward trend is obvious — and this is the 0.99 correlation from Correlation and Covariance, made visible.

A scatter plot shows the shape of a relationship, which a correlation coefficient can't. Here the shape is a straight line, so 0.99 is a fair summary. When it isn't a line, the number misleads and only the plot tells you.

Useful arguments: s= for point size, c= for colour, alpha= for transparency when points overlap.

---

Pie chart

department_df = pd.DataFrame({
    "Department": ["CSE", "IT", "ECE"],
    "Students": [4, 3, 3],
})

plt.figure()
plt.pie(department_df["Students"],
        labels=department_df["Department"],
        autopct="%1.1f%%")
plt.title("Students by Department")
plt.show()

labels= names the slices; autopct="%1.1f%%" writes the percentage on each — the format string means "one decimal place, then a literal %" (the %% escapes it).

Note the counts sum to 10, the whole dataset, so "share of total" is a meaningful question here. That's the condition a pie chart needs.

Still, three slices at 40/30/30 are hard to rank by eye. A bar chart of the same counts would be easier to read — see why pie charts are usually wrong.

---

Several charts in one figure

plt.subplots() returns a figure and an array of axes:

fig, axes = plt.subplots(1, 2, figsize=(10, 3.8))

axes[0].bar(df["Student"], df["Marks"])
axes[0].set_title("Marks")
axes[0].tick_params(axis="x", rotation=45)

axes[1].scatter(df["Study_Hours"], df["Marks"])
axes[1].set_title("Hours vs Marks")

plt.tight_layout()
plt.show()

Two things change when you use axes directly:

• Methods live on the axis, not on plt — axes[0].bar(...) rather than plt.bar(...)
• Labelling methods gain a set_ prefix — set_title(), set_xlabel(), set_ylabel()

plt.tight_layout() stops labels overlapping. rotation=45 makes the ten student names readable.

Two interfaces, one library

plt.title() is the pyplot interface — it acts on whatever figure is "current". ax.set_title() is the object-oriented interface, where you say explicitly which axis you mean.

Pyplot is shorter for one chart. The object-oriented style is what you need for subplots, and it's what seaborn returns, so both are worth recognising.

---

Saving to a file

plt.show() opens a window. To write an image instead:

plt.savefig("chart.png", dpi=150, bbox_inches="tight")

Argument	Does
dpi=150	resolution — 150+ for anything printed or embedded
bbox_inches="tight"	trims surrounding whitespace and stops labels being cut off

warning

savefig() must come before show()

show() clears the figure, so calling savefig() afterwards writes a blank image. This is a frequent and confusing bug.

Also, in a script with no display, add matplotlib.use("Agg") before importing pyplot — otherwise matplotlib may fail trying to open a window. Every chart on this page was rendered that way.

---

Useful extras

plt.figure(figsize=(10, 4))     # size in inches
plt.grid(True)                  # gridlines
plt.legend()                    # needs label= on each plot call
plt.xticks(rotation=45)         # rotate tick labels
plt.ylim(0, 100)                # fix the axis range
plt.style.use("seaborn-v0_8")   # a nicer default look

Passing label= to a plot call and then calling plt.legend() is how you identify multiple series on one chart. Without the label=, the legend is empty.

---

Common Mistakes

#	Mistake	Result	Fix
1	Forgetting plt.figure()	charts drawn on top of each other	start each with figure()
2	savefig() after show()	blank image	save first
3	Line plot over categories	implies a false progression	use bar()
4	No axis labels	unreadable chart	xlabel(), ylabel(), title()
5	legend() with no label=	empty legend	plt.plot(..., label="x")
6	plt.title() on a subplot	lands on the wrong axis	ax.set_title()
7	Default dpi for embedding	blurry	dpi=150 or higher
8	No bbox_inches="tight"	labels cut off	pass it to savefig
9	hist() on a categorical column	meaningless	bar() with counts

---

Summary

Chart	Call
Line	plt.plot(x, y, marker="o")
Bar	plt.bar(x, y) / plt.barh(y, x)
Histogram	plt.hist(col, bins=n)
Scatter	plt.scatter(x, y)
Pie	plt.pie(vals, labels=..., autopct="%1.1f%%")
Grid of charts	fig, axes = plt.subplots(r, c)
Save	plt.savefig(f, dpi=150, bbox_inches="tight")

Key takeaways

• The pattern is always: figure() → draw → label → show()
• plt.figure() prevents charts merging into one another
• savefig() before show(), or you get a blank file
• plt.* acts on the current figure; ax.set_* is explicit and needed for subplots
• Bars compare categories, histograms show one column's distribution — not interchangeable
• The bin count changes a histogram's apparent shape; choose it deliberately
• A scatter plot shows the shape a correlation coefficient can't
• Label every axis; an unlabelled chart isn't evidence of anything

See also: Choosing a Chart · Seaborn for the same charts in less code · Plotly for interactive versions

---

Run It Yourself

matplotlib_charts.py

import pandas as pd
import matplotlib
matplotlib.use("Agg")          # no display needed; remove for an interactive window
import matplotlib.pyplot as plt

df = pd.DataFrame({
    "Student": ["Arun", "Bala", "Charan", "Divya", "Esha",
                "Farah", "Gokul", "Hari", "Indhu", "Jaya"],
    "Study_Hours": [2, 3, 5, 1, 4, 3, 4, 6, 2, 5],
    "Attendance": [70, 75, 90, 60, 85, 78, 88, 95, 72, 92],
    "Marks": [55, 62, 85, 45, 78, 68, 80, 95, 58, 88],
})

# bar — the right chart for comparing students
plt.figure(figsize=(7, 4))
plt.bar(df["Student"], df["Marks"])
plt.title("Marks of Students")
plt.xlabel("Student")
plt.ylabel("Marks")
plt.xticks(rotation=45)
plt.savefig("bar.png", dpi=150, bbox_inches="tight")   # BEFORE show()
plt.close()

# histogram — distribution of one column
plt.figure(figsize=(7, 4))
plt.hist(df["Marks"], bins=5, edgecolor="white")
plt.title("Distribution of Marks")
plt.xlabel("Marks")
plt.ylabel("Number of Students")
plt.savefig("hist.png", dpi=150, bbox_inches="tight")
plt.close()

# scatter — relationship between two columns
plt.figure(figsize=(7, 4))
plt.scatter(df["Study_Hours"], df["Marks"], s=80, alpha=0.8)
plt.title("Study Hours vs Marks")
plt.xlabel("Study Hours")
plt.ylabel("Marks")
plt.grid(True, alpha=0.3)
plt.savefig("scatter.png", dpi=150, bbox_inches="tight")
plt.close()

print("wrote bar.png, hist.png, scatter.png")
print(f"correlation: {df['Study_Hours'].corr(df['Marks']):.4f}")

Output

wrote bar.png, hist.png, scatter.png
correlation: 0.9895

---

Practice Questions

From the Unit 2 question bank. Tags and marks are explained on the Python index.

Unit 2 § K — Matplotlib

K1. [THEORY] What is Matplotlib, and what is pyplot? Why is it imported as plt? [3]

K2. [PLOT] For df_viz, draw a line plot of Student against Marks with circle markers, a title, and both axis labels. [4]

K3. [PLOT] Add grid lines to K2, then customise them to be dashed, width 0.7, and semi-transparent. [3]

K4. [PLOT] Draw a bar plot of Student against Marks, with the x-tick labels rotated 45°. [4]

K5. [PLOT] Draw the same chart as a horizontal bar plot. Name the function that changes. [2]

K6. [PLOT] Draw a scatter plot of Study_Hours against Marks with a title, axis labels and a grid. [3]

K7. [PLOT] Draw a histogram of Marks with 5 bins and black bar edges. [3]

K8. [THEORY] A classic viva question. What is the difference between a bar plot and a histogram? What is on the x-axis of each, and what kind of data does each need? [4]

K9. [PLOT] Draw a pie chart of students per department, showing percentages to one decimal place. Name the parameter that produces those percentages. [4]

K10. [THEORY] What is an exploded pie chart, and which parameter creates it? What is a donut chart, and how do you make one? [4]

K11. [PLOT] Draw a box plot of Marks. Name the five values it displays and say how outliers appear. [4]

K12. [PLOT] Draw multiple line plots on one set of axes — Study_Hours, Attendance and Marks against Student — with a legend. [5]

K13. [PLOT] Draw a stacked bar chart of Study_Hours and Attendance per student. Name the parameter that stacks the second series. [4]

K14. [THEORY] What does plt.figure() do, and why does every plot in your Data visualization with matplotlib.py start with it? What happens if you leave it out? [3]

K15. [THEORY] What does plt.show() do? Why does the next plot start empty after you call it? [3]

K16. [PROG] Name the function for each: chart title · x-axis label · y-axis label · grid · legend · x-tick rotation · figure size · saving to a file. [4]

Unit 2 § O — EDA Mini Challenges

O3. [PROG] (Lab 5.) Write one script that draws all six of these for df_viz: line, bar, scatter, histogram, box plot (Seaborn), and an annotated correlation heatmap — each with a title and axis labels. [10]

O6. [PROG] For df_viz, build a single figure holding four subplots in a 2×2 grid — line, bar, scatter, histogram — each titled. Name the function that makes the grid. [6]