Artificial Intelligence Course for Young Learners in Tampines, Singapore

1. Introduction: Accessing and Modifying DataFrame Subsets

Once a dataset is loaded into a DataFrame, data scientists rarely process the entire table at once. Machine learning preprocessing, exploratory analysis, and data cleaning require accessing, filtering, and modifying specific subsets of data.

Pandas provides two primary methods for subset selection:

1. loc (Location) — label-based selection
2. iloc (Index Location) — integer position-based selection

• df.loc[row_labels, col_labels] — label-based; slicing includes both start and end boundaries
• df.iloc[row_positions, col_positions] — integer positions; slicing excludes the end boundary
• Boolean indexing: df[df['Column'] > value]
• Conditional assignment: df.loc[condition, col] = value

2. Environment Setup: Creating the Sample DataFrame

Import Pandas as pd and build a DataFrame with four features (Name, Age, City, Salary) and four records:

import pandas as pd

# Creating a DataFrame with 4 features and 4 records
data = {
    "Name": ["Alice", "Bob", "Charlie", "David"],
    "Age": [25, 35, 28, 42],
    "City": ["New York", "London", "Paris", "Tokyo"],
    "Salary": [70000, 85000, 62000, 95000]
}

df = pd.DataFrame(data)
print(df)

Output structure (default row indices 0–3):

      Name  Age      City  Salary
0    Alice   25  New York   70000
1      Bob   35    London   85000
2  Charlie   28     Paris   62000
3    David   42     Tokyo   95000

3. Label-Based Selection: df.loc

The df.loc indexer selects data using row and column labels.

Syntax: df.loc[row_selection, column_selection].

# Selecting rows with labels 1 to 3, and columns 'Name' and 'Salary'
subset_loc = df.loc[1:3, ["Name", "Salary"]]
print(subset_loc)

      Name  Salary
1      Bob   85000
2  Charlie   62000
3    David   95000

• Row labels here match default indices (0, 1, 2, 3).
• Columns are selected by name: "Name" and "Salary".
• Critical rule for loc: both endpoints are included — 1:3 returns labels 1, 2, and 3 (three rows).

4. Position-Based Selection: df.iloc

The df.iloc (Integer Location) indexer selects strictly by zero-based integer positions for rows and columns.

Syntax: df.iloc[row_positions, column_positions].

# Selecting row indices 0:2 and column indices 1:3
subset_iloc = df.iloc[0:2, 1:3]
print(subset_iloc)

   Age      City
0   25  New York
1   35    London

• Rows 0:2 → indices 0 and 1 (stop 2 excluded).
• Columns 1:3 → Age and City (column 3 Salary excluded).

5. Key Difference: Boundary Inclusivity (loc vs. iloc)

Indexer	Based On	Start	End	Example (1:3)
df.loc	Labels / names	Included	Included	Labels 1, 2, 3 (3 records)
df.iloc	Integer positions	Included	Excluded	Positions 1 and 2 (2 records)

In iloc, the endpoint is excluded (standard Python slicing). In loc, both endpoints are included because labels are identifiers, not offsets.

6. Boolean Indexing: Filtering Rows with Conditions

# Filtering records where Salary is greater than 75,000
high_salary = df[df["Salary"] > 75000]
print(high_salary)

    Name  Age    City  Salary
1    Bob   35  London   85000
3  David   42   Tokyo   95000

Only Bob (85,000) and David (95,000) satisfy the condition.

7. Modifying Data Subsets Using df.loc

Using column labels with loc is preferred over integer positions: large tables make counting columns error-prone, and reordering or adding columns would break positional code.

Example — set salary to 65,000 for employees under age 30:

# Targeting rows where Age < 30 and updating the 'Salary' column
df.loc[df["Age"] < 30, "Salary"] = 65000

# Inspecting records where Age < 30 to verify the change
print(df[df["Age"] < 30])

      Name  Age      City  Salary
0    Alice   25  New York   65000
2  Charlie   28     Paris   65000

Alice and Charlie are updated; the filtered print confirms both now show 65000.