🐍 Python Data Science

Practicals — Line-by-Line Study Guide

Practicals 1 • 2 • 3 | Easy Learn & Memorize Edition

📚 How to Use This Guide

This document breaks every line of code into simple English. Use the following method to memorize
fast:

• Read the CODE column — understand what it says.

• Read the EXPLANATION column — understand WHY it is written.
• Cover the code and try to rewrite it from memory.
• Focus on the MEMORY TRICK boxes — they give you shortcuts to recall code quickly.

📊 Practical 1
Data Cleaning & Correlation Analysis using Seaborn

Practical 1 — Data Cleaning & Correlation

Goal: Load a dataset, find missing values, remove duplicates, detect outliers using IQR method, and
compute correlations between numeric columns.

Step 1 — Import Libraries

💻 Code 📖 What it does / Why it's used
import seaborn as sns Imports Seaborn (data visualization library). 'as sns' means
you can type sns instead of seaborn — shorter!
import pandas as pd Imports Pandas (data table/spreadsheet library). 'as pd' is
💻 Code 📖 What it does / Why it's used
the standard short name.
import [Link] as Imports Matplotlib's plotting module. All charts are shown
plt using this.
import numpy as np Imports NumPy (math and array library). Used for
percentile calculations.

🧠 Memory Trick — Import Order

Always remember: S-P-M-N = Seaborn, Pandas, Matplotlib, NumPy. Think: 'Some People
Make Numbers'

Step 2 — Load & Inspect Data

💻 Code 📖 What it does / Why it's used
df = pd.read_csv('[Link]') Reads the CSV file named '[Link]' from your folder and
stores it in a variable called df (DataFrame = a table).
print(df) Displays the full table in the console.
dfnomis = [Link]() Counts non-null (non-missing) values in each column.
Stores result in dfnomis.
dftotal = [Link][0] Gets total number of rows. shape gives (rows, columns), so
shape[0] = rows.
df_mis = dftotal - dfnomis Subtracts non-missing count from total rows = gives you
missing values per column.
print("Missing values...\n", Prints the missing value count for each column with a label.
df_mis)

🧠 Memory Trick — Missing Values Formula

Missing = Total Rows - Non-Null Count → df_mis = dftotal - dfnomis

Step 3 — Clean Data (Duplicates & Nulls)

💻 Code 📖 What it does / Why it's used
df1 = df.drop_duplicates() Removes duplicate rows (rows that are exactly the same).
Result stored in df1.
df1 = [Link]() Removes rows that have any missing (NaN) value. Keeps
only complete rows.
print(df1) Shows the cleaned table.
 💡 Why clean data?
Machine learning and statistics only work correctly on clean, complete data. Dirty data =
wrong results.

Step 4 — Outlier Detection using IQR Method

IQR = Inter-Quartile Range. It measures the 'middle spread' of data. Values too far outside this range
are called outliers.

💻 Code 📖 What it does / Why it's used

Q1 = [Link](df1['sal'], Gets the 25th percentile (Q1) of the 'sal' column — the
25) value below which 25% of data falls.
Q3 = [Link](df1['sal'], Gets the 75th percentile (Q3) — the value below which
75) 75% of data falls.
IQR = Q3 - Q1 Calculates the Inter-Quartile Range = spread of the middle
50% of data.
ul = Q3 + 1.5 * IQR Upper Limit: anything above this is an outlier (too high).
ll = Q1 - 1.5 * IQR Lower Limit: anything below this is an outlier (too low).
df1_non_outliers = Filters the DataFrame — keeps only rows where salary is
df1[(df1['sal'] >= ll) & within the limits (no outliers).
(df1['sal'] <= ul)]
print("Length with outliers:", Shows how many rows existed before removing outliers.
len(df1))
print("Length without Shows how many rows remain after removing outliers.
outliers:",
len(df1_non_outliers))

🧠 IQR Formula to Memorize

Lower Limit = Q1 - 1.5×IQR | Upper Limit = Q3 + 1.5×IQR Think: 'Stretch 1.5 times the
middle range on both sides.'

Step 5 — Correlation Analysis

💻 Code 📖 What it does / Why it's used
print(df['exper'].corr(df['sal Pearson: measures linear relationship between experience
'], method='pearson')) & salary. Value between -1 and 1.
print(df['exper'].corr(df['sal Spearman: rank-based correlation. Good when data is not
'], method='spearman')) normally distributed.
print(df['exper'].corr(df['sal Kendall: another rank-based method. More robust with
'], method='kendall')) small samples or many ties.
df_corr = Creates a full correlation matrix for all numeric columns in
 💻 Code 📖 What it does / Why it's used
[Link](numeric_only=True) the dataset.
print(df_corr) Prints the correlation matrix table.
print(df['sal'].median()) Shows the median salary value (middle value when
sorted).
print([Link]()) Prints summary statistics: count, mean, std, min, max, Q1,
Q3 for all numeric columns.

🧠 3 Correlation Methods — Quick Recall

P-S-K = Pearson (linear), Spearman (rank), Kendall (robust rank) Think: 'Please Send
Knowledge'

Step 6 — Heatmap Visualization

💻 Code 📖 What it does / Why it's used
[Link](df_corr, vmin=-1, Draws a colored grid (heatmap) of the correlation matrix.
vmax=1, annot=True) annot=True shows numbers inside each cell. vmin/vmax
set the color scale from -1 to 1.
[Link]() Displays the chart on screen.

💡 Why use a Heatmap?

A heatmap makes it easy to spot strong correlations (dark red/blue) vs weak ones (light
colors) at a glance.

🌸 Practical 2
Iris Dataset — Exploration & Inspection

Practical 2 — Iris Dataset Exploration

Goal: Load the famous Iris dataset, inspect its shape, data types, missing values, and basic statistics.

Step 1 — Import Libraries

Same imports as Practical 1: seaborn, pandas, matplotlib, numpy. No new imports needed.
Step 2 — Load & Basic Info
💻 Code 📖 What it does / Why it's used
df = pd.read_csv('[Link]') Loads the [Link] file into a DataFrame called df.
print(df) Prints the entire dataset.
print([Link]) Prints total number of elements = rows × columns.
print("Rows = ", [Link][0]) Prints only the number of rows.
print("Cols = ", [Link][1]) Prints only the number of columns.
print([Link]) Prints (rows, columns) together as a tuple.
print([Link]) Prints the raw data as a NumPy array (no column
headers).

🧠 shape vs size
shape = (rows, cols) → structure | size = rows × cols → total cells Think: Shape =
skeleton, Size = total bricks.

Step 3 — Tail, Head & Summary

💻 Code 📖 What it does / Why it's used
print([Link]()) Prints the last 5 rows of the dataset. Useful to see if data
ends correctly.
print([Link]()) Prints the first 5 rows. Most commonly used to get a quick
look.
print([Link]()) Prints statistics: count, mean, std dev, min, 25%, 50%,
75%, max for each numeric column.
print([Link]()) Prints column names, data types, and non-null counts.
Shows the structure of the data.
print([Link]().sum()) Counts missing (NaN) values per column. isna() returns
True/False; sum() counts the Trues.

🧠 head vs tail
head() = first 5 rows (top of table) | tail() = last 5 rows (bottom of table) Think: head = face,
tail = end.

Step 4 — Clean Data

💻 Code 📖 What it does / Why it's used
df1 = [Link]() Drops all rows with any missing values. Stores clean data
in df1.
 💻 Code 📖 What it does / Why it's used
print(df.drop_duplicates(subse Shows only unique species names — removes duplicate
t=['species'])) species rows. subset means only check that column for
duplicates.

🚢 Practical 3
Titanic Dataset — Visualization & Analysis

Practical 3 — Titanic Visualization

Goal: Load the Titanic dataset, clean it, and create 4 types of visualizations: Count Plot, Scatter Plot,
KDE/Density Plot, and Pair Plot.

Step 1 — Import Libraries

💻 Code 📖 What it does / Why it's used
import pandas as pd Pandas for data tables.
import seaborn as sns Seaborn for statistical charts.
import [Link] as Matplotlib for displaying charts.
plt

Step 2 — Load & Clean Data

💻 Code 📖 What it does / Why it's used
data = Loads the built-in Titanic dataset from Seaborn. No CSV
sns.load_dataset("titanic") file needed!
data = Removes rows where age, fare, sex, or survived is
[Link](subset=['age', missing. Only keeps rows that have all 4 values.
'fare', 'sex', 'survived'])

💡 Why subset in dropna?

Without subset, dropna removes rows with ANY missing value in any column. subset limits it
to only the columns you care about — so you don't lose too much data.
Chart 1 — Count Plot (Survival by Gender)
💻 Code 📖 What it does / Why it's used
[Link]() Creates a new blank figure/canvas for the chart.
[Link](x='sex', Draws a bar chart counting male/female passengers, split
hue='survived', data=data) by survived (0=No, 1=Yes). hue adds color-coding by
survived.
[Link]("Survival Count by Adds a title at the top of the chart.
Gender")
[Link]("Gender") Labels the X-axis.
[Link]("Count") Labels the Y-axis.
[Link](title='Survived', Adds a legend showing which color means survived/not
labels=['No', 'Yes']) survived.
[Link]() Displays the chart.

🧠 countplot keyword
countplot → counts occurrences and makes bars hue → splits bars by a second category
(adds color groups)

Chart 2 — Scatter Plot (Age vs Fare)

💻 Code 📖 What it does / Why it's used
[Link]() New blank canvas.
[Link](x='age', Plots each passenger as a dot: X = age, Y = fare paid.
y='fare', hue='survived', Color shows if they survived.
data=data)
[Link]("Scatter Plot: Age Chart title.
vs Fare")
[Link]("Age") X-axis label.
[Link]("Fare") Y-axis label.
[Link](title='Survived') Adds legend for survived colors.
[Link]() Displays the chart.

💡 What does a Scatter Plot show?

Each dot = one person. You can see if older passengers paid more, or if fare affected
survival. Look for patterns or clusters of dots.
Chart 3 — KDE / Density Plot (Age & Fare Distribution)
💻 Code 📖 What it does / Why it's used
[Link]() New blank canvas.
[Link](data=data['age'], Draws a smooth density curve for Age. fill=True shades the
label='Age', fill=True) area under the curve. Shows where most passengers'
ages were concentrated.
[Link](data=data['fare'], Same thing for Fare — shows how fare amounts were
label='Fare', fill=True) distributed. Both curves appear on the same chart.
[Link]("Density Chart title.
Distribution of Age and Fare")
[Link]("Values") X-axis label.
[Link]("Density") Y-axis label — 'density' means probability, not count.
[Link]() Adds a legend to distinguish Age curve from Fare curve.
[Link]() Shows the chart.

🧠 KDE = Kernel Density Estimate

Think of KDE as a smoothed histogram. Instead of bars, it draws a smooth curve showing
where values cluster. fill=True makes it look like a mountain with colored area.

Chart 4 — Pair Plot (All Relationships at Once)

💻 Code 📖 What it does / Why it's used
[Link](data[['age', Creates a grid of scatter plots for every combination of age,
'fare', 'pclass', fare, pclass, survived. Diagonal shows distribution of each
'survived']], hue='survived') variable. hue colors points by survived. Automatically calls
[Link]() internally.
[Link]() Ensures the pair plot is displayed.

🧠 Pair Plot — What to look for?

Diagonal = histogram of each variable Off-diagonal = scatter between two variables Colors =
survived (blue) vs not survived (orange)

⚡ Quick Cheat Sheet

Most important functions across all 3 practicals
Quick Reference Cheat Sheet

💻 Code 📖 What it does / Why it's used

pd.read_csv('[Link]') Load CSV file into DataFrame
[Link] Returns (rows, cols)
[Link]() Count non-null values per column
df.drop_duplicates() Remove duplicate rows
[Link]() Remove rows with missing values
[Link]().sum() Count missing values per column
[Link]() Summary statistics (mean, std, min, max...)
[Link]() Column types and null info
[Link]() / [Link]() First/Last 5 rows
[Link](col, 25) Get Q1 (25th percentile)
[Link](col, 75) Get Q3 (75th percentile)
[Link](col2, Pearson correlation between two columns
method='pearson')
[Link](numeric_only=True) Full correlation matrix
[Link](df_corr, Draw heatmap of correlation matrix
annot=True)
[Link](x=..., hue=...) Count bar chart, colored by category
[Link](x=..., y=..., Scatter plot colored by category
hue=...)
[Link](data=..., Smooth density curve
fill=True)
[Link](df, hue=...) Grid of scatter plots for all column pairs
[Link]() Create a new blank chart
[Link]() Display the chart
[Link]() / [Link]() / Add title and axis labels
[Link]()
[Link]() Show the legend

✅ Study Tip: Write out the code once by hand — muscle memory is the best way to memorize
code!