Mahmoud Abas, Full-Stack Software Engineer & AI

Machine Learning (2026-2027) - Data Visualization with Matplotlib and Seaborn

In the print dialog, choose "Save as PDF" as the destination.

Machine Learning, Week 2

Data Visualization with Matplotlib and Seaborn

Look at the data before you model it: the charts every notebook of the course uses, and when each one fits.

Objectives

Draw scatter, bar, histogram, line and box plots with Matplotlib, with titles, labels, ticks and legends
Compute histogram bins and the parts of a boxplot by hand
Plot straight from a DataFrame with df.plot.scatter and df.plot.hist
Compare groups with Seaborn: boxplot, violinplot, regplot, lmplot
Compute a correlation r and read a correlation heatmap
Choose the chart that fits the question, and use it in your proposal

Week 2 of the plan

Where This Sits in the Course

Notebook: Matplotlib_Seaborn_basics, on three small datasets
Real-world scenario: the plan leaves it open this week, so we use the notebook's own: clinic patients, a month of polls, 178 wines
Project milestone this week: Project/Paper Proposal

Plan for the Two Hours

Part	What we do	Time
1	Matplotlib on the clinic data, a histogram by hand	30 min
2	The poll data: line plot, a boxplot by hand	20 min
3	Pandas and Seaborn on the wine data	20 min
4	Correlation, the heatmap, choosing a chart	15 min
5	Practice with answers, then takeaways	35 min

Open the Notebook in Colab

text

https://colab.research.google.com/github/
tirthajyoti/Machine-Learning-with-Python/blob/master/
Pandas%20and%20Numpy/Matplotlib_Seaborn_basics.ipynb

Join the three lines into one address, or click "Open in Colab" on the lesson page.

Part 1

Matplotlib on the Clinic Data

One figure, built call by call

The Clinic Data

Twelve patients: age in years, weight in kg, height in cm

python

people = ['P1', 'P2', 'P3', 'P4', 'P5', 'P6',
          'P7', 'P8', 'P9', 'P10', 'P11', 'P12']
age    = [21, 12, 32, 45, 37, 18,
          28, 52, 5, 40, 48, 15]
weight = [55, 35, 77, 68, 70, 60,
          72, 69, 18, 65, 82, 48]
height = [160, 135, 170, 165, 173, 168,
          175, 159, 105, 171, 155, 158]

Plain Python lists, one value per patient, in the same order
Matplotlib accepts lists, NumPy arrays and DataFrame columns alike
Question: is there a relationship between these features?

A First Scatter Plot

One dot per patient: age on x, height on y

python

import matplotlib.pyplot as plt

plt.scatter(age, height)
plt.show()

Scatter plot of age against height for the 12 patients

Bells and Whistles

Each extra call adds one element to the same figure

Call	Adds
`plt.figure(figsize=(8, 6))`	figure size in inches
`plt.title`, `plt.xlabel`, `plt.ylabel`	the texts around the plot
`plt.ylim(100, 200)`	the visible y range
`plt.xticks(list)`	where x tick marks go

Call	Adds
`plt.scatter(c=, s=, edgecolors=)`	dot colour, size, outline
`plt.text(x, y, s)`	text at a data point
`plt.vlines`, `plt.hlines`	vertical, horizontal lines
`plt.legend(list, loc=2)`	a legend, upper left

Worked Example: Which Ticks?

python

plt.xticks([i * 5 for i in range(12)],
           fontsize=15)

text

[0, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55]

range(12) gives i = 0 to 11, times 5: twelve ticks up to 55.

The styled scatter plot with title, grid, text, dashed lines and a legend

Bar Chart

One number across named items; every bar starts at zero

python

plt.figure(figsize=(12, 4))
plt.bar(x=people, height=weight,
        width=0.6, color='orange',
        edgecolor='k', alpha=0.6)
plt.xticks(fontsize=14, rotation=30)
plt.show()

width=0.6: each bar fills 60% of its slot
alpha=0.6: partly transparent
rotation=30: tilted labels

Step 1 of 3

Worked Example: Histogram Bins

python

plt.hist(weight, color='red', edgecolor='k', alpha=0.75, bins=5)

w = 82 − 185 = 645 = 12.8

Step 2 of 3

Step 2: Count the Values in Each Bin

A bin holds its left edge, not its right one; the last bin also holds 82

Bin	Weights inside	Count
18 to 30.8	18	1
30.8 to 43.6	35	1
43.6 to 56.4	55, 48	2
56.4 to 69.2	68, 60, 69, 65	4
69.2 to 82	77, 70, 72, 82	4

Step 3 of 3

Step 3: Check Against the Plot

Counts 1, 1, 2, 4, 4 add up to 12, one per patient
plt.hist returns the same counts and edges
The two short bars on the left are the two children

Part 2

The Poll Data

A line through time, and a box around a spread

A Month of Polls, Drawn as Lines

python

np.random.seed(0)   # added
days = np.arange(1, 31)
candidate_A = 50 + days * 0.07 \
              + 2 * np.random.randn(30)
candidate_B = 50 - days * 0.1 \
              + 3 * np.random.randn(30)

plt.style.use('fivethirtyeight')
plt.plot(days, candidate_A, 'o-', c='blue')
plt.plot(days, candidate_B, '^-', c='green')

A: +0.07 a day, B: -0.1 a day, plus random noise
plt.style.use stays on for every later plot

Line plot of both candidates over 30 days

The Parts of a Boxplot

Median: the middle of the sorted values, the line in the box
Q1 and Q3: the quartiles; the box holds the middle half
IQR = Q3 - Q1, the length of the box
Fences at 1.5 IQR beyond the box decide what is unusual
Whiskers reach the farthest values inside the fences; outliers beyond them are circles

Step 1 of 5

Worked Example: Box of the 12 Heights

Sort, then find the median

text

sorted:   105 135 155 158 159 160 165 168 170 171 173 175
position:   0   1   2   3   4   5   6   7   8   9  10  11

p = 0.5 (n − 1) = 0.5 × 11 = 5.5

Q₂ = 160 + 0.5 × (165 − 160) = 162.5

Step 2 of 5

Step 2: The Quartiles Q1 and Q3

Q₁ = 155 + 0.75 × (158 − 155) = 157.25

Q₃ = 170 + 0.25 × (171 − 170) = 170.25

Step 3 of 5

Step 3: IQR and the Fences

Q₃ − Q₁ = 170.25 − 157.25 = 13

Q₁ − 1.5 × 13 = 137.75 Q₃ + 1.5 × 13 = 189.75

Step 4 of 5

Step 4: Whiskers and Outliers

Fences: 137.75 and 189.75

Part	Value	Why
Lower whisker	155	smallest height at or above 137.75
Upper whisker	175	largest height at or below 189.75
Outliers	105, 135	below the lower fence: P9 and P2, the children

Step 5 of 5

Step 5: The Mean, and the Check

python

plt.boxplot(x=[height], showmeans=True)

μ = 189412 ≈ 157.83

Boxplot of the 12 heights with two outliers and the mean as a triangle

Try It: Build the Boxplot Step by Step

The Notebook's Boxplot of the Polls

python

plt.style.use('ggplot')
plt.boxplot(x=[candidate_A, candidate_B],
            showmeans=True)
plt.xticks([1, 2], ['Candidate A', 'Candidate B'])

	A	B
Median	51.947	47.389
Q1 to Q3	50.87 to 53.53	45.35 to 48.89
Outlier	46.364	54.452
Mean	51.971	47.581

Side-by-side boxplots of the two candidates

Part 3

Pandas and Seaborn on the Wine Data

Name the columns, let the library do the rest

Load the Wine Data

python

df = pd.read_csv("https://raw.githubusercontent.com/tirthajyoti/"
                 "Stats_data_science_ValleyML/master/Notebooks/Data/wine.data.csv")
df.head()

df.shape is (178, 14): 178 wines, 14 columns
Class takes the values 1, 2, 3 (59, 71 and 48 wines)
13 measurements, such as Alcohol, Flavanoids, Color intensity, Proline

Pandas Plots Straight from a DataFrame

python

df.plot.scatter('Alcohol', 'Color intensity')
plt.show()

Name the columns; no lists to pass
Pandas calls Matplotlib for you, so plt.xlabel and friends still work
Each dot is one wine

Pandas scatter plot of Alcohol against Color intensity

Worked Example: 20 Bins of Alcohol

python

df['Alcohol'].plot.hist(bins=20, figsize=(5, 5),
                         edgecolor='k')

w = 14.83 − 11.0320 = 3.8020 = 0.19

Seaborn: One Box per Class

python

import seaborn as sns

sns.boxplot(x='Class', y='Alcohol', data=df)

Seaborn boxplots of Alcohol for each Class

Violin Plot

python

sns.violinplot(x='Class', y='Alcohol', data=df)

Violin plot: A boxplot plus the shape of the distribution: the width at each height shows how many wines have that value.

Seaborn violin plots of Alcohol for each Class

regplot: A Scatter Plot with a Fitted Line

python

sns.regplot(x='Alcohol', y='Color intensity',
            data=df)

The line that fits the dots best: a linear regression, next week's topic
The shaded band is the confidence interval: how unsure the line is
Fitted here: slope 1.560, intercept -15.226

Seaborn regplot of Alcohol against Color intensity with a regression line

Worked Example: Read the Line at Alcohol 13

ŷ = 1.560 x − 15.226

ŷ = 1.560 × 13 − 15.226 ≈ 5.05

lmplot: One Line per Group

python

sns.lmplot(x='Alcohol', y='Color intensity',
           hue='Class', data=df)

hue='Class' colours the dots by class and fits one line per class

Class	Slope
1	1.094
2	0.464
3	1.527

Seaborn lmplot with one regression line per Class

Part 4

Correlation and the Heatmap

Every pair of columns in one picture

The Correlation Coefficient r

r = ∑ (x − μ_x)(y − μ_y)√(∑ (x − μ_x)² × ∑ (y − μ_y)²)

r near 1: high values of x go with high values of y
r near -1: high x goes with low y
r near 0: no straight-line relation

Step 1 of 2

Worked Example: r for Four Patients

P1, P4, P6, P10: heights 160, 165, 168, 171 and weights 55, 68, 60, 65

Patient	Deviations	Product	Squares
P1	-6, -7	42	36, 49
P4	-1, 6	-6	1, 36
P6	2, -2	-4	4, 4
P10	5, 3	15	25, 9
Sum		47	66, 98

Step 2 of 2

Step 2: Combine into r

r = 47√(66 × 98) = 4780.424 ≈ 0.584

The Correlation Matrix and Its Heatmap

python

corr_mat = np.corrcoef(df, rowvar=False)
corr_mat.shape      # (14, 14)
corr_df = pd.DataFrame(corr_mat,
    columns=df.columns, index=df.columns)
sns.heatmap(corr_df, linewidth=1,
            cmap='plasma')

rowvar=False: the variables are the columns
14 x 14 = 196 values of r
Diagonal is 1; the matrix is symmetric
Yellow: r near 1; dark blue: most negative

Heatmap of the 14 by 14 correlation matrix of the wine data

Worked Example: Read the Heatmap

Question	Answer	r
Most related to `Class`	`Flavanoids`	-0.847
Next two	`OD280/OD315`, `Total phenols`	-0.788, -0.719
Most related pair	`Total phenols` and `Flavanoids`	0.865
Least related pair	`Ash` and `OD280/OD315`	0.004

Choosing a Chart

Question	Chart
Do two numbers move together?	scatter, `regplot`
How does a number change over time?	line
How do named items compare?	bar

Question	Chart
What shape does one column have?	histogram
How do groups differ in spread?	box, violin
Which of many columns are related?	heatmap

Try It: The Chart Chooser

Common Mistakes

A line plot through rows that have no order, such as the 12 patients
Judging a histogram with one bin count only
Forgetting that plt.style.use stays on for later plots
Deleting boxplot outliers without checking them: they are only values beyond 1.5 IQR
Reading r = 0 as "no relation": r only sees straight-line relations
Treating a correlation as a cause

Project Milestone: Project/Paper Proposal

State the idea, or the paper you will reproduce, in two or three sentences
Name the dataset: source, rows, columns, and the target column
Draw a histogram of the target, or a bar chart of its classes
Draw a boxplot or scatter of the target against the column you expect to matter most
Draw the correlation heatmap and note the strongest relations
Put the plots in the proposal, with one sentence under each

Practice

Practice: Your Turn

About 30 minutes

About 7 minutes

Practice 1: A Histogram by Hand

w = 44 − 124 = 8

Bin	Values	Count
12 to 20	12, 15	2
20 to 28	21, 22, 26	3
28 to 36	30, 31, 33	3
36 to 44	38, 44	2

About 8 minutes

Practice 2: A Boxplot by Hand

Part	Position	Value
Q1	0.25 x 9 = 2.25	55 + 0.25 x 2 = 55.5
Median	0.5 x 9 = 4.5	58 + 0.5 x 2 = 59
Q3	0.75 x 9 = 6.75	61 + 0.75 x 2 = 62.5

About 5 minutes

Practice 3: r by Hand

∑ (x − μ_x)(y − μ_y) = 4.5 + 0.5 + 0 + 6 = 11

r = 11√(5 × 26) = 1111.402 ≈ 0.965

About 10 minutes

Practice 4: In Colab, on the Wine Data

sns.boxplot(x='Class', y='Flavanoids', data=df): which class has the highest median? Which boxes show outliers?
df['Proline'].plot.hist(bins=10): how many wines in the tallest bin, between which edges? (np.histogram gives exact numbers)
sns.regplot(x='Total phenols', y='Flavanoids', data=df): slope sign? r with np.corrcoef?
Heatmap of ['Class', 'Alcohol', 'Flavanoids', 'Proline'] only: which pair has the largest |r|?

Practice 4: Answer

Task	Answer
1	Medians 2.98, 2.03, 0.685: class 1. Outliers: class 2 (5.08), class 3 (1.57)
2	41 wines between 558.4 and 698.6
3	Positive: slope 1.380, r = 0.865
4	Class and Flavanoids, r = -0.847, the same as in the full heatmap

Key Takeaways

Matplotlib builds a figure call by call: data, then title, labels, ticks, text, lines, legend
A histogram cuts the range into equal bins; the bin count changes the picture
A boxplot shows median, quartiles, whiskers to 1.5 IQR, and outliers beyond
Pandas and Seaborn plot DataFrame columns by name; x= or hue= splits by group
r measures straight-line relation; a heatmap shows every pair at once
Pick the chart from the question: time, compare, shape, spread, relation

Open this lesson

Mahmoud AbasData Visualization with Matplotlib and Seaborn

Part

What we do

Time

Matplotlib on the clinic data, a histogram by hand

30 min

The poll data: line plot, a boxplot by hand

20 min

Pandas and Seaborn on the wine data

20 min

Correlation, the heatmap, choosing a chart

15 min

Practice with answers, then takeaways

35 min

people = ['P1', 'P2', 'P3', 'P4', 'P5', 'P6', 'P7', 'P8', 'P9', 'P10', 'P11', 'P12'] age = [21, 12, 32, 45, 37, 18, 28, 52, 5, 40, 48, 15] weight = [55, 35, 77, 68, 70, 60, 72, 69, 18, 65, 82, 48] height = [160, 135, 170, 165, 173, 168, 175, 159, 105, 171, 155, 158]

Call

Adds

plt.figure(figsize=(8, 6))

figure size in inches

plt.title, plt.xlabel, plt.ylabel

the texts around the plot

plt.ylim(100, 200)

the visible y range

plt.xticks(list)

where x tick marks go

Call

Adds

plt.scatter(c=, s=, edgecolors=)

dot colour, size, outline

plt.text(x, y, s)

text at a data point

plt.vlines, plt.hlines

vertical, horizontal lines

plt.legend(list, loc=2)

a legend, upper left

Bin

Weights inside

Count

18 to 30.8

30.8 to 43.6

43.6 to 56.4

55, 48

56.4 to 69.2

68, 60, 69, 65

69.2 to 82

77, 70, 72, 82

np.random.seed(0) # added days = np.arange(1, 31) candidate_A = 50 + days * 0.07 \ + 2 * np.random.randn(30) candidate_B = 50 - days * 0.1 \ + 3 * np.random.randn(30) plt.style.use('fivethirtyeight') plt.plot(days, candidate_A, 'o-', c='blue') plt.plot(days, candidate_B, '^-', c='green')

Part

Value

Why

Lower whisker

155

smallest height at or above 137.75

Upper whisker

175

largest height at or below 189.75

Outliers

105, 135

below the lower fence: P9 and P2, the children

Median

51.947

47.389

Q1 to Q3

50.87 to 53.53

45.35 to 48.89

Outlier

46.364

54.452

Mean

51.971

47.581

Class

Slope

1.094

0.464

1.527

Patient

Deviations

Product

Squares

-6, -7

36, 49

-1, 6

-6

1, 36

2, -2

-4

4, 4

P10

5, 3

25, 9

Sum

66, 98

corr_mat = np.corrcoef(df, rowvar=False) corr_mat.shape # (14, 14) corr_df = pd.DataFrame(corr_mat, columns=df.columns, index=df.columns) sns.heatmap(corr_df, linewidth=1, cmap='plasma')

Question

Answer

Most related to Class

Flavanoids

-0.847

Next two

OD280/OD315, Total phenols

-0.788, -0.719

Most related pair

Total phenols and Flavanoids

0.865

Least related pair

Ash and OD280/OD315

0.004

Question

Chart

Do two numbers move together?

scatter, regplot

How does a number change over time?

line

How do named items compare?

bar

Question

Chart

What shape does one column have?

histogram

How do groups differ in spread?

box, violin

Which of many columns are related?

heatmap

Bin

Values

Count

12 to 20

12, 15

20 to 28

21, 22, 26

28 to 36

30, 31, 33

36 to 44

38, 44

Part

Position

Value

0.25 x 9 = 2.25

55 + 0.25 x 2 = 55.5

Median

0.5 x 9 = 4.5

58 + 0.5 x 2 = 59

0.75 x 9 = 6.75

61 + 0.75 x 2 = 62.5

Task

Answer

Medians 2.98, 2.03, 0.685: class 1. Outliers: class 2 (5.08), class 3 (1.57)

41 wines between 558.4 and 698.6

Positive: slope 1.380, r = 0.865

Class and Flavanoids, r = -0.847, the same as in the full heatmap