EDA and Feature Engineering Test

Section A: Multiple Choice Questions (12 x 0.5 = 6 marks)

Which imputation method is most suitable for skewed numerical data?
a. Mean
b. Median
c. Mode
d. Zero
dropna(axis=1) removes:
a. Rows with missing values
b. Columns with missing values
c. Only numeric missing values
d. Only categorical missing values
One-hot encoding converts categories into:
a. Single integer
b. Binary columns
c. Probability values
d. Ranks
Which encoding can cause dimensionality explosion?
a. Ordinal
b. Target
c. One-hot
d. Label
Which plot best shows data distribution?
a. Line plot
b. Bar chart
c. Histogram
d. Heatmap
A correlation value of 0 indicates:
a. Strong positive relation
b. Strong negative relation
c. No linear relation
d. Perfect relation
Which method ensures every data point appears in test set exactly once?
a. Hold-out
b. Random split
c. K-Fold
d. Stratified split
Normalization scales data between:
a. -1 and 1
b. 0 and 1
c. Mean 0, std 1
d. Any range
Which scaler is sensitive to outliers?
a. StandardScaler
b. MinMaxScaler
c. RobustScaler
d. MaxAbsScaler
Feature engineering improves:
a. Data size
b. Model performance
c. Data leakage
d. Missing values
Suppose you have a column “Chest Pain Type” which features [‘Normal’, ‘Moderate’, and, ‘Severe’, what is the best way to encode it?
a. One Hot Encoding
b. Nominal Encoding
c. Ordinal Encoding
d. Frequency Based Encoding
Which distribution does the following histogram is shown here:

right-skewed-histogram
a. Left Skew
b. Right Skew
c. Normal
d. Uniform

Section B: Debug the Code (6 × 1 = 6 Marks)

1. This code results in an Error. Why? Fix it.

df.fillna(mean)  

Debugged:

df.fillna(df['column_name'].mean())
# df.fillna(df.mean(), inplace=True)

2. From this code snippet, debug the error:

import pandas as pd  
import matplotlib.pyplot as plt  
  
### i. Loading the Dataset  
df = pd.read_csv("adult.csv")  
  
### ii. Cheking the null values

df.isna.sum()  
  
### iii. Checking the Bar Plot for Gender column  
gender_counts = df['sex'].value_counts()  
  
# Extract categories and their counts  
categories = gender_counts.index.tolist()  
counts = gender_counts.values.tolist()  
  
# Plot using matplotlib  
plt.figure(figsize=(6,4))  
plt.plot(categories, counts)  
plt.title('Count of Each Gender')
plt.xlabel('Gender')  
plt.ylabel('Count')  
  
### iv. Encode the Column Male with 1 and Female with 0:  
  
df ['gender'] = df['gender'].rename({"Male":1, "Female":0})  

Debugged:

import pandas as pd
import matplotlib.pyplot as plt

# i. Loading the Dataset
df = pd.read_csv("adult.csv")

# ii. Checking the null values
df.isna().sum() # 🕷️df.isna.sum() is wrong

# iii. Checking the Bar Plot for Gender column
gender_counts = df['sex'].value_counts()

# Extract categories and their counts
categories = gender_counts.index.tolist()
counts = gender_counts.values.tolist()

# Plot using matplotlib
plt.figure(figsize=(6, 4))
plt.bar(categories, counts) # 🕷️ Using plt.plot() for categorical data, it's not appropriate
plt.title('Count of Each Gender')
plt.xlabel('Gender')
plt.ylabel('Count')
plt.show()

# iv. Encode the Column Male with 1 and Female with 0
# 🕷️ Encoding using .rename(): .rename() is for renaming labels
# 🕷️ Encoding wrong column: df['gender'] does not exist
df['gender'] = df['sex'].map({'Male': 1, 'Female': 0})

Errors and Fix

Error	Fix
`df.isna.sum()`	Missing `()` → `df.isna().sum()`
`plt.plot(categories, counts)`	Not suitable for categorical data → use `plt.bar(categories, counts)`
`.rename()` used for encoding	`.rename()` is for renaming labels → use `.map()`
Encoding on `df['gender']`	Column does not exist → encode from `df['sex']`
`df['sex']` (spacing issue in original)	Remove extra space → `df['sex']`

3. This code results in an Error. Why? Fix it.

OneHotEncoder.fit_transform(data)

Debugged:

from sklearn.preprocessing import OneHotEncoder

encoder = OneHotEncoder(sparse_output=False)
encoded_data = encoder.fit_transform(data)

4. Why does unpacking throw an error?

from sklearn.model_selection import train_test_split
X_train, y_train, X_test, y_test = train_test_split(X, y, test_size=0.2)

Debugged:

from sklearn.model_selection import train_test_split

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state = 42, shuffle=True)

5. This code results in an Error. Why? Fix it.

plt.hist(df)

Debugged:

plt.hist(df['numerical_column'])
plt.show()

6. This code results in an Error. Why? Fix it.

df.interpolate(method='average')

Debugged:

df.interpolate(method='linear')

Section D: Coding Questions (2 × 2.5 = 5 Marks)

1. Create a new feature BMI from Weight and Height, then select top 3 features using correlation.

Example dataset:

Weight	Height	Age	BP
60	1.65	22	120
70	1.70	25	130
80	1.75	30	140
90	1.80	35	150

\[{BMI} = \frac{\text{Weight (in kilograms)}}{\text{Height}^2 \text{ (in meters)}}\]

Explanation

BMI formula: Weight / Height²
Correlation: df.corr() calculates pairwise correlation.
Top features: Sorted by absolute correlation with BMI.
- index[1:4] skips BMI itself.

import pandas as pd

# Create DataFrame
data = {
    'Weight': [60, 70, 80, 90],
    'Height': [1.65, 1.70, 1.75, 1.80],
    'Age': [22, 25, 30, 35],
    'BP': [120, 130, 140, 150]
}

df = pd.DataFrame(data)

# Create BMI feature
df['BMI'] = df['Weight'] / (df['Height'] ** 2)

# Check correlation
corr = df.corr()

# Select top 3 features correlated with BMI
top_features = corr['BMI'].abs().sort_values(ascending=False).index[1:4]

print("DataFrame with BMI:")
print(df)
print("\nTop 3 features correlated with BMI:", list(top_features))

Expected Output:

   Weight  Height  Age   BP        BMI
0      60    1.65   22  120  22.038567
1      70    1.70   25  130  24.221453
2      80    1.75   30  140  26.122449
3      90    1.80   35  150  27.777778

Top 3 features correlated with BMI: ['Weight', 'BP', 'Height']

2. Given a dataset with a Date column, split it into:

Year
Month
Day

Date	Sales
2024-01-15	1000
2024-02-20	1500
2024-03-10	1200

import pandas as pd

# Create DataFrame
df = pd.DataFrame({
    'Date': ['2024-01-15', '2024-02-20', '2024-03-10'],
    'Sales': [1000, 1500, 1200]
})

# Convert Date to datetime and split into Year, Month, Day
df['Date'] = pd.to_datetime(df['Date'])
df['Year'] = df['Date'].dt.year
df['Month'] = df['Date'].dt.month
df['Day'] = df['Date'].dt.day

print(df)

Expected Output:

        Date  Sales  Year  Month  Day
2024-01-15   1000  2024      1   15
2024-02-20   1500  2024      2   20
2024-03-10   1200  2024      3   10