Practice Exercises: Pandas Basics¶

The only way to get good at pandas is to write pandas code. Reading notes builds recognition; writing code builds fluency. These exercises are structured in three levels. Work through warm-up before main, and main before stretch. Try each exercise without looking at the solution first.

Setup¶

import pandas as pd
import numpy as np

Warm-Up Exercises¶

These exercises check that you can use the core operations without hesitation.

Warm-Up 1 — Build and Inspect a DataFrame¶

Create the following student grades DataFrame and answer the questions.

grades = pd.DataFrame({
    "student_id": ["S001", "S002", "S003", "S004", "S005", "S006"],
    "name": ["Priya", "Rohan", "Amit", "Divya", "Karan", "Neha"],
    "section": ["A", "B", "A", "B", "A", "B"],
    "math_score": [88, 74, 92, 65, 79, 85],
    "english_score": [76, 81, 70, 88, 72, 90],
    "attendance_pct": [95.0, 82.0, 98.0, 74.0, 88.0, 91.0]
})

Tasks:

Print the first 3 rows.
Print the shape.
Print the dtypes.
Select only the name column.
Select name, math_score, and english_score.
Add a column total_score that is the sum of math_score and english_score.
Add a column average_score that is the average of the two scores.
Rename attendance_pct to attendance.

Show answer

# 1. First 3 rows
print(grades.head(3))
# Output:
#   student_id   name section  math_score  english_score  attendance_pct
# 0       S001  Priya       A          88             76            95.0
# 1       S002  Rohan       B          74             81            82.0
# 2       S003   Amit       A          92             70            98.0

# 2. Shape
print(grades.shape)
# Output: (6, 6)

# 3. Data types
print(grades.dtypes)
# Output:
# student_id        object
# name              object
# section           object
# math_score         int64
# english_score      int64
# attendance_pct   float64

# 4. Single column (returns Series)
print(grades["name"])

# 5. Multiple columns (returns DataFrame)
print(grades[["name", "math_score", "english_score"]])

# 6. Total score
grades["total_score"] = grades["math_score"] + grades["english_score"]

# 7. Average score
grades["average_score"] = grades["total_score"] / 2

# 8. Rename column
grades = grades.rename(columns={"attendance_pct": "attendance"})

print(grades.head())

Warm-Up 2 — .loc and .iloc Practice¶

Using the grades DataFrame from Warm-Up 1:

Use .loc to select the row with label 2 (Amit's row).
Use .iloc to select the third row (same row, using position).
Use .loc to select rows 0 through 3, columns name and math_score.
Use .iloc to select the first 3 rows and the first 2 columns.
Use .loc to select all rows where section == "A", keeping only name and average_score.

Show answer

# 1. Row with label 2 using .loc
print(grades.loc[2])
# Output: student_id    S003, name    Amit, ...

# 2. Third row using .iloc (position 2 = third row, 0-indexed)
print(grades.iloc[2])

# 3. Rows 0-3, specific columns with .loc (inclusive on both ends)
print(grades.loc[0:3, ["name", "math_score"]])
# Output:
#     name  math_score
# 0  Priya          88
# 1  Rohan          74
# 2   Amit          92
# 3  Divya          65

# 4. First 3 rows, first 2 columns with .iloc (end excluded)
print(grades.iloc[0:3, 0:2])
# Output:
#   student_id   name
# 0       S001  Priya
# 1       S002  Rohan
# 2       S003   Amit

# 5. Filter + column selection with .loc
section_a = grades.loc[grades["section"] == "A", ["name", "average_score"]]
print(section_a)
# Output:
#     name  average_score
# 0  Priya           82.0
# 2   Amit           81.0
# 4  Karan           75.5

Warm-Up 3 — Basic Filtering¶

Using the grades DataFrame:

Filter students with math_score above 80.
Filter students in Section B.
Filter students with attendance below 85.
Filter students in Section A AND with math_score above 85.
Filter students in Section A OR with english_score above 85.
Find the top 2 students by math_score.

Show answer

# 1. Math score above 80
print(grades[grades["math_score"] > 80])

# 2. Section B
print(grades[grades["section"] == "B"])

# 3. Attendance below 85
print(grades[grades["attendance"] < 85])

# 4. Section A AND high math score
high_math_a = grades[
    (grades["section"] == "A") &
    (grades["math_score"] > 85)
]
print(high_math_a[["name", "section", "math_score"]])
# Output:
#     name section  math_score
# 0  Priya       A          88
# 2   Amit       A          92

# 5. Section A OR high English score
result = grades[
    (grades["section"] == "A") |
    (grades["english_score"] > 85)
]
print(result[["name", "section", "english_score"]])

# 6. Top 2 by math score
print(grades.nlargest(2, "math_score")[["name", "math_score"]])
# Output:
#     name  math_score
# 2   Amit          92
# 0  Priya          88

Main Exercises¶

These exercises require combining multiple operations and thinking through the right sequence.

Main 1 — Sales Data Analysis¶

Work with this e-commerce dataset:

orders = pd.DataFrame({
    "order_id": [f"ORD{1000+i}" for i in range(1, 16)],
    "customer_id": ["C001", "C002", "C003", "C001", "C004", "C002", "C005",
                    "C003", "C001", "C004", "C005", "C002", "C003", "C001", "C004"],
    "product": ["Laptop", "Headphones", "Mouse", "Keyboard", "Laptop", "Monitor",
                "Headphones", "Laptop", "Mouse", "Keyboard", "Monitor", "Mouse",
                "Headphones", "Keyboard", "Laptop"],
    "category": ["Electronics", "Electronics", "Accessories", "Accessories", "Electronics",
                 "Electronics", "Electronics", "Electronics", "Accessories", "Accessories",
                 "Electronics", "Accessories", "Electronics", "Accessories", "Electronics"],
    "units_sold": [1, 2, 5, 3, 1, 1, 2, 2, 8, 4, 1, 6, 1, 2, 1],
    "unit_price": [74999, 3499, 599, 1299, 74999, 16999, 3499, 74999, 599, 1299,
                   16999, 599, 3499, 1299, 74999],
    "city": ["Delhi", "Mumbai", "Bangalore", "Delhi", "Pune", "Mumbai", "Bangalore",
             "Delhi", "Mumbai", "Bangalore", "Pune", "Delhi", "Mumbai", "Pune", "Delhi"],
    "order_month": [1, 1, 1, 2, 2, 2, 3, 3, 3, 1, 1, 2, 2, 3, 3]
})

Tasks:

Add a revenue column (units_sold * unit_price).
Print the total revenue across all orders.
Print the average revenue per order.
Print the number of orders per category.
Find the 3 largest orders by revenue.
Filter orders where revenue > 50000 and city is in ["Delhi", "Mumbai"].
Find the total revenue per city, sorted descending.
Find the total revenue per month.
Find how many unique customers placed orders.
Find which customer placed the most orders.

Show answer

# 1. Revenue column
orders["revenue"] = orders["units_sold"] * orders["unit_price"]

# 2. Total revenue
print(f"Total revenue: ₹{orders['revenue'].sum():,}")
# Output: Total revenue: ₹589,469

# 3. Average revenue per order
print(f"Average order revenue: ₹{orders['revenue'].mean():,.0f}")
# Output: Average order revenue: ₹39,298

# 4. Orders per category
print(orders["category"].value_counts())
# Output:
# Electronics    9
# Accessories    6

# 5. Top 3 orders by revenue
top_3 = orders.nlargest(3, "revenue")[["order_id", "product", "city", "revenue"]]
print(top_3)

# 6. High-value orders in target cities
high_value_metro = orders[
    (orders["revenue"] > 50000) &
    (orders["city"].isin(["Delhi", "Mumbai"]))
]
print(high_value_metro[["order_id", "customer_id", "product", "city", "revenue"]])

# 7. Revenue by city (sorted descending)
city_revenue = orders.groupby("city")["revenue"].sum().sort_values(ascending=False)
print(city_revenue)

# 8. Revenue by month
monthly_revenue = orders.groupby("order_month")["revenue"].sum()
print(monthly_revenue)

# 9. Unique customer count
print(f"Unique customers: {orders['customer_id'].nunique()}")
# Output: Unique customers: 5

# 10. Customer with most orders
top_customer = orders["customer_id"].value_counts().head(1)
print(top_customer)
# Output: C001 = 4 orders

Main 2 — Reading a CSV with Messy Properties¶

Create this CSV content (save it as employees_messy.csv or simulate with io.StringIO):

import io

csv_content = """Employee ID,Full Name,Department,Monthly Salary,Join Date,Performance
E001,Priya Sharma,Sales,52000,2022-03-15,Good
E002,Rohan Mehta,Engineering,88000,2019-07-01,Excellent
E003,Amit Patel,Engineering,79000,2020-11-20,Good
E004,Divya Singh,HR,46000,2023-01-10,Average
E005,Karan Rao,Sales,67000,2021-05-25,Excellent
E006,Neha Gupta,Engineering,92000,2018-09-12,Excellent
E007,Suresh Kumar,HR,44000,2023-06-01,Good
E008,Anjali Desai,Sales,55000,2022-08-19,Average
"""

employees = pd.read_csv(
    io.StringIO(csv_content),
    parse_dates=["Join Date"]
)

Tasks:

Print .info() — note the column names.
Rename columns: "Employee ID" → "emp_id", "Full Name" → "name", "Monthly Salary" → "monthly_salary", "Join Date" → "join_date", "Performance" → "performance", "Department" → "department".
Sort employees by monthly_salary descending and print the top 3.
Filter employees in "Engineering" with monthly_salary > 80000.
Find the average salary per department.
Count how many employees are in each performance category.
Select only name and monthly_salary for employees who joined after 2021-01-01.
Add a column annual_salary equal to monthly_salary * 12.

Show answer

# 1. Info check
employees.info()

# 2. Rename columns
employees = employees.rename(columns={
    "Employee ID": "emp_id",
    "Full Name": "name",
    "Department": "department",
    "Monthly Salary": "monthly_salary",
    "Join Date": "join_date",
    "Performance": "performance"
})

# 3. Top 3 by salary
print(employees.nlargest(3, "monthly_salary")[["name", "department", "monthly_salary"]])
# Output:
#          name  department  monthly_salary
# 5  Neha Gupta Engineering           92000
# 1  Rohan Mehta Engineering           88000
# 2   Amit Patel Engineering           79000

# 4. Engineering with salary > 80000
eng_high = employees[
    (employees["department"] == "Engineering") &
    (employees["monthly_salary"] > 80000)
]
print(eng_high[["name", "monthly_salary"]])

# 5. Average salary per department
dept_avg = employees.groupby("department")["monthly_salary"].mean().round(0)
print(dept_avg)

# 6. Performance distribution
print(employees["performance"].value_counts())

# 7. Recent joiners — name and salary only
recent_joiners = employees.loc[
    employees["join_date"] > "2021-01-01",
    ["name", "monthly_salary"]
]
print(recent_joiners)

# 8. Annual salary
employees["annual_salary"] = employees["monthly_salary"] * 12
print(employees[["name", "monthly_salary", "annual_salary"]].head())

Main 3 — Missing Value Detection and Handling¶

customer_survey = pd.DataFrame({
    "customer_id": ["C001", "C002", "C003", "C004", "C005", "C006", "C007", "C008"],
    "age": [28, None, 34, 45, None, 31, 27, 39],
    "city": ["Delhi", "Mumbai", None, "Pune", "Bangalore", None, "Delhi", "Mumbai"],
    "purchase_amount": [4500.0, 12000.0, 8900.0, None, 3200.0, 6700.0, None, 9100.0],
    "satisfaction_score": [4, 5, None, 3, 4, None, 5, 4],
    "membership_tier": ["Silver", "Gold", "Silver", None, "Bronze", "Gold", "Bronze", None]
})

Tasks:

Print missing value counts per column.
Print the percentage missing per column.
Print all rows that have at least one missing value.
Calculate the average purchase_amount (pandas skips NaN automatically — verify this).
Fill missing city values with "Unknown".
Fill missing age with the median age.
Fill missing satisfaction_score with the mode (most common value).
Drop rows where purchase_amount is missing.
After all fills and drops, verify there are no more missing values.

Show answer

# 1. Missing counts
print(customer_survey.isna().sum())
# Output:
# customer_id          0
# age                  2
# city                 2
# purchase_amount      2
# satisfaction_score   2
# membership_tier      2

# 2. Percentage missing
missing_pct = customer_survey.isna().mean() * 100
print(missing_pct.round(1))

# 3. Rows with any missing value
rows_with_na = customer_survey[customer_survey.isna().any(axis=1)]
print(rows_with_na)

# 4. Average purchase_amount (NaN skipped automatically)
avg_purchase = customer_survey["purchase_amount"].mean()
print(f"Average: ₹{avg_purchase:,.0f}")
print(f"Count used in mean: {customer_survey['purchase_amount'].count()} of {len(customer_survey)}")

# Work on a copy to avoid SettingWithCopyWarning
survey = customer_survey.copy()

# 5. Fill missing city
survey["city"] = survey["city"].fillna("Unknown")

# 6. Fill missing age with median
age_median = survey["age"].median()
survey["age"] = survey["age"].fillna(age_median)
print(f"Age median used: {age_median}")

# 7. Fill missing satisfaction_score with mode
satisfaction_mode = survey["satisfaction_score"].mode()[0]
survey["satisfaction_score"] = survey["satisfaction_score"].fillna(satisfaction_mode)

# 8. Drop rows where purchase_amount is still missing
survey = survey.dropna(subset=["purchase_amount"])

# 9. Verify — should show all zeros
print(survey.isna().sum())

Stretch Exercises¶

These exercises require combining multiple techniques. There is no single correct solution — focus on getting correct output.

Stretch 1 — Full Sales Analysis Pipeline¶

store_sales = pd.DataFrame({
    "transaction_id": [f"TXN{1000+i}" for i in range(1, 21)],
    "date": pd.date_range("2024-01-01", periods=20, freq="3D"),
    "store_id": ["S01", "S02", "S01", "S03", "S02", "S01", "S03", "S02", "S01", "S03",
                 "S02", "S01", "S03", "S02", "S01", "S03", "S02", "S01", "S03", "S01"],
    "product_name": ["Rice 5kg", "Cooking Oil 1L", "Detergent 2kg", "Rice 5kg", "Sugar 1kg",
                     "Cooking Oil 1L", "Detergent 2kg", "Rice 5kg", "Sugar 1kg", "Cooking Oil 1L",
                     "Rice 5kg", "Detergent 2kg", "Sugar 1kg", "Cooking Oil 1L", "Rice 5kg",
                     "Detergent 2kg", "Sugar 1kg", "Cooking Oil 1L", "Rice 5kg", "Detergent 2kg"],
    "units": [10, 5, 8, 12, 20, 6, 15, 11, 25, 4, 9, 10, 18, 7, 14, 12, 22, 5, 16, 9],
    "unit_price": [250, 180, 320, 250, 85, 180, 320, 250, 85, 180,
                   250, 320, 85, 180, 250, 320, 85, 180, 250, 320],
    "store_city": ["Delhi", "Mumbai", "Delhi", "Bangalore", "Mumbai", "Delhi", "Bangalore",
                   "Mumbai", "Delhi", "Bangalore", "Mumbai", "Delhi", "Bangalore", "Mumbai",
                   "Delhi", "Bangalore", "Mumbai", "Delhi", "Bangalore", "Delhi"]
})

Tasks:

Add a revenue column.
Print a complete first-look inspection (shape, dtypes, missing values, describe).
Find total revenue per store (store_id).
Find which store has the highest average revenue per transaction.
Find the top 3 products by total revenue across all stores.
Filter transactions where units > 15 and print store and product details.
Find which city generates the most total revenue.
Add a high_volume column: True if units > 12, False otherwise.
Sort the DataFrame by date ascending and by revenue descending within the same date.
Print a summary: total transactions, total revenue, unique products, unique stores.

Show answer

# 1. Revenue column
store_sales["revenue"] = store_sales["units"] * store_sales["unit_price"]

# 2. First-look inspection
print(store_sales.shape)
print(store_sales.dtypes)
print(store_sales.isna().sum())
print(store_sales.describe())

# 3. Revenue per store
store_revenue = store_sales.groupby("store_id")["revenue"].sum().sort_values(ascending=False)
print(store_revenue)

# 4. Highest average revenue per transaction by store
store_avg = store_sales.groupby("store_id")["revenue"].mean().sort_values(ascending=False)
print(f"Best store by average transaction value: {store_avg.index[0]}")

# 5. Top 3 products by total revenue
product_revenue = store_sales.groupby("product_name")["revenue"].sum().nlargest(3)
print(product_revenue)

# 6. High-unit transactions
high_unit = store_sales[store_sales["units"] > 15][["store_id", "product_name", "units", "revenue"]]
print(high_unit)

# 7. Revenue by city
city_revenue = store_sales.groupby("store_city")["revenue"].sum().sort_values(ascending=False)
print(f"Top city: {city_revenue.index[0]} with ₹{city_revenue.iloc[0]:,}")

# 8. High-volume flag
store_sales["high_volume"] = np.where(store_sales["units"] > 12, True, False)

# 9. Multi-column sort
store_sales_sorted = store_sales.sort_values(
    ["date", "revenue"],
    ascending=[True, False]
)
print(store_sales_sorted[["date", "store_id", "product_name", "revenue"]].head(8))

# 10. Summary
print(f"Total transactions: {len(store_sales)}")
print(f"Total revenue: ₹{store_sales['revenue'].sum():,}")
print(f"Unique products: {store_sales['product_name'].nunique()}")
print(f"Unique stores: {store_sales['store_id'].nunique()}")

Stretch 2 — SettingWithCopyWarning Challenge¶

This exercise is about understanding views vs. copies — the most confusing pandas behavior for beginners.

inventory = pd.DataFrame({
    "item_code": ["A01", "A02", "A03", "B01", "B02", "C01"],
    "item_name": ["Widget A", "Widget B", "Gadget C", "Tool D", "Tool E", "Part F"],
    "category": ["Widgets", "Widgets", "Gadgets", "Tools", "Tools", "Parts"],
    "stock_qty": [50, 0, 120, 30, 0, 80],
    "unit_cost": [250, 180, 490, 320, 140, 95]
})

Tasks:

Filter rows where stock_qty == 0 and try to add a column reorder_flag = True. Observe whether you get a SettingWithCopyWarning.
Fix the issue using .copy().
Fix the same operation using .loc directly on the original DataFrame instead.
Explain in a comment when you would choose .copy() vs .loc.

Show answer

# Task 1: This triggers SettingWithCopyWarning in most pandas versions
out_of_stock = inventory[inventory["stock_qty"] == 0]
out_of_stock["reorder_flag"] = True   # Warning here — may not affect original
# The modification may be silently lost

# Task 2: Fix with .copy() — you own the copy outright
out_of_stock = inventory[inventory["stock_qty"] == 0].copy()
out_of_stock["reorder_flag"] = True   # Safe — no warning
print(out_of_stock)
# Output shows reorder_flag = True for the zero-stock items

# Task 3: Fix using .loc on the original
inventory["reorder_flag"] = False   # Start with default value for all rows
inventory.loc[inventory["stock_qty"] == 0, "reorder_flag"] = True
print(inventory[["item_name", "stock_qty", "reorder_flag"]])
# Output:
#   item_name  stock_qty  reorder_flag
# 0  Widget A         50         False
# 1  Widget B          0          True
# 2  Gadget C        120         False
# 3    Tool D         30         False
# 4    Tool E          0          True
# 5    Part F         80         False

# Task 4: When to use which approach:
# Use .copy() when you need an independent subset to work with separately
#   (e.g., you are building a report on just out-of-stock items)
# Use .loc when you want to modify the original DataFrame in place
#   (e.g., adding a flag column back into the main dataset)

Self-Check¶

Before moving to Day 03 (Pandas Advanced), confirm you can do all of these without looking at notes:

[ ] Create a DataFrame from a dictionary and from a list of dicts
[ ] Use .head(), .shape, .dtypes, .info(), and .isna().sum() immediately after loading
[ ] Select a single column (returning a Series) and multiple columns (returning a DataFrame)
[ ] Use .loc[row_condition, column_list] for combined row filtering and column selection
[ ] Use .iloc[row_positions, col_positions] for position-based access
[ ] Filter with &, |, ~ and know why and/or do not work
[ ] Use .isin() for membership filtering
[ ] Use .query() for readable multi-condition filters
[ ] Sort by one and multiple columns with mixed ascending/descending
[ ] Use .nlargest() and .nsmallest() for top/bottom N
[ ] Add calculated columns using vectorized expressions
[ ] Use .value_counts(), .nunique(), .describe(), and .corr()
[ ] Explain what SettingWithCopyWarning means and how to avoid it
[ ] Save a DataFrame to CSV with index=False

Previous: Basic Data Analysis | Next: Interview Questions