Encoding categorical variables in Data Analysis Python - Time & Space Complexity

Time Complexity: Encoding categorical variables

O(n x k)

Understanding Time Complexity

When we convert categories into numbers, we want to know how long it takes as data grows.

How does the time needed change when we have more data or more categories?

Scenario Under Consideration

Analyze the time complexity of the following code snippet.

from sklearn.preprocessing import OneHotEncoder
import pandas as pd

data = pd.DataFrame({
    'color': ['red', 'blue', 'green', 'blue', 'red'] * 1000
})

encoder = OneHotEncoder(sparse_output=False)
encoded = encoder.fit_transform(data[['color']])

This code turns a list of colors into a set of columns with 1s and 0s for each color.

Identify Repeating Operations

Identify the loops, recursion, array traversals that repeat.

• Primary operation: Scanning each row to assign a 1 or 0 for each category.
• How many times: Once for each data row times the number of unique categories.

How Execution Grows With Input

As the number of rows or categories grows, the work grows too.

Input Size (n rows)	Approx. Operations
10	10 x number of categories
100	100 x number of categories
1000	1000 x number of categories

Pattern observation: The work grows in a straight line with the number of rows and categories.

Final Time Complexity

Time Complexity: O(n x k)

This means the time grows with both the number of data rows and the number of unique categories.

Common Mistake

[X] Wrong: "Encoding time depends only on the number of rows."

[OK] Correct: Because each row must be checked against all categories, more categories mean more work.

Interview Connect

Understanding how encoding scales helps you explain data preparation steps clearly and shows you think about efficiency.

Self-Check

"What if we used label encoding instead of one-hot encoding? How would the time complexity change?"

Practice

1. What is the main purpose of encoding categorical variables in data science?

easy

A. To visualize data using charts and graphs

B. To remove missing values from the dataset

C. To convert categories into numbers so algorithms can process them

D. To normalize numerical data between 0 and 1

Solution

1. Step 1: Understand the role of encoding

   Encoding changes categories like 'red', 'blue' into numbers so computers can understand them.

2. Step 2: Identify the purpose in data science

   Machine learning models require numbers, so encoding helps prepare data for these models.

3. Final Answer:

   To convert categories into numbers so algorithms can process them -> Option C

4. Quick Check:

   Encoding = convert categories to numbers [OK]

Hint: Encoding means turning words into numbers for models [OK]

Common Mistakes:

• Confusing encoding with data cleaning
• Thinking encoding normalizes data
• Assuming encoding is for visualization

2. Which of the following is the correct way to apply label encoding using scikit-learn in Python?

easy

A. from sklearn.preprocessing import LabelEncoder
le = LabelEncoder()
data['color_encoded'] = le.fit_transform(data['color'])

B. from sklearn.preprocessing import OneHotEncoder
le = OneHotEncoder()
data['color_encoded'] = le.fit_transform(data['color'])

C. import pandas as pd
data['color_encoded'] = pd.get_dummies(data['color'])

D. data['color_encoded'] = data['color'].astype('category').cat.codes

Solution

1. Step 1: Recognize label encoding syntax

   LabelEncoder from sklearn.preprocessing is used with fit_transform on a column.

2. Step 2: Match code to correct usage

   from sklearn.preprocessing import LabelEncoder le = LabelEncoder() data['color_encoded'] = le.fit_transform(data['color']) correctly imports LabelEncoder, creates an instance, and applies fit_transform to the 'color' column.

3. Final Answer:

   from sklearn.preprocessing import LabelEncoder le = LabelEncoder() data['color_encoded'] = le.fit_transform(data['color']) -> Option A

4. Quick Check:

   LabelEncoder usage = from sklearn.preprocessing import LabelEncoder le = LabelEncoder() data['color_encoded'] = le.fit_transform(data['color']) [OK]

Hint: LabelEncoder uses fit_transform on a single column [OK]

Common Mistakes:

• Using OneHotEncoder instead of LabelEncoder
• Trying to assign get_dummies output directly to one column
• Forgetting to import LabelEncoder

3. Given the following code, what will be the output of print(encoded_df)?

import pandas as pd
from sklearn.preprocessing import OneHotEncoder

data = pd.DataFrame({'fruit': ['apple', 'banana', 'apple', 'orange']})
encoder = OneHotEncoder(sparse=False)
encoded_array = encoder.fit_transform(data[['fruit']])
encoded_df = pd.DataFrame(encoded_array, columns=encoder.categories_[0])
print(encoded_df)

medium

A. An error because OneHotEncoder cannot handle string data

B. A DataFrame with columns 'apple', 'banana', 'orange' containing 1s and 0s indicating fruit presence

C. A DataFrame with a single column containing encoded numbers like 0, 1, 2

D. A DataFrame with original 'fruit' column and encoded columns combined

Solution

1. Step 1: Understand OneHotEncoder output

   OneHotEncoder creates separate columns for each category with 1 or 0 indicating presence.

2. Step 2: Check code behavior

   Encoded array has shape (4,3) for 'apple', 'banana', 'orange'. DataFrame columns are set to these categories.

3. Final Answer:

   A DataFrame with columns 'apple', 'banana', 'orange' containing 1s and 0s indicating fruit presence -> Option B

4. Quick Check:

   OneHotEncoder output = separate 0/1 columns [OK]

Hint: OneHotEncoder creates one column per category with 0/1 [OK]

Common Mistakes:

• Expecting a single column with numbers
• Thinking OneHotEncoder fails on strings
• Assuming original column stays in encoded_df

4. Identify the error in this code snippet for label encoding and how to fix it:

from sklearn.preprocessing import LabelEncoder
data = pd.DataFrame({'color': ['red', 'blue', 'green']})
le = LabelEncoder()
encoded = le.fit_transform(data['color'])
print(encoded)

medium

A. data is a dict, not a DataFrame; convert to DataFrame before encoding

B. LabelEncoder cannot encode colors; use OneHotEncoder instead

C. fit_transform should be called on le, not on data['color']

D. Missing import for pandas; add 'import pandas as pd'

Solution

1. Step 1: Check imports and dependencies

   The code uses pd.DataFrame without importing pandas, causing NameError: name 'pd' is not defined.

2. Step 2: Fix by adding the import

   Add 'import pandas as pd' at the beginning, before defining data.

3. Final Answer:

   Missing import for pandas; add 'import pandas as pd' -> Option D

4. Quick Check:

   Missing pandas import for pd.DataFrame [OK]

Hint: Always import pandas before using pd.DataFrame [OK]

Common Mistakes:

• Not importing pandas before using pd.DataFrame
• Confusing LabelEncoder with OneHotEncoder
• Ignoring data structure requirements

5. You have a dataset with a column 'size' containing categories: 'small', 'medium', 'large', and 'medium'. You want to encode it so that 'small' = 0, 'medium' = 1, and 'large' = 2, but also keep track of the original order. Which encoding method and approach should you use?

hard

A. Use pandas Categorical with ordered=True and then convert to codes

B. Use OneHotEncoder to create separate columns for each size

C. Use LabelEncoder directly without ordering

D. Manually replace categories with numbers using a dictionary without ordering

Solution

1. Step 1: Understand the need to preserve order

   Since 'small' < 'medium' < 'large', encoding must respect this order.

2. Step 2: Use pandas Categorical with ordered=True

   Creating an ordered categorical allows converting categories to codes that reflect the order.

3. Step 3: Convert ordered categorical to codes

   Codes will be 0 for 'small', 1 for 'medium', 2 for 'large' as desired.

4. Final Answer:

   Use pandas Categorical with ordered=True and then convert to codes -> Option A

5. Quick Check:

   Ordered categorical preserves order in encoding [OK]

Hint: Use ordered pandas Categorical to keep category order [OK]

Common Mistakes:

• Using LabelEncoder which ignores order
• Using OneHotEncoder which loses order info
• Replacing categories without defining order