Normalization (standardize, min-max) quiz

Question 1

What is feature normalization in machine learning, and why is it often needed?

Accepted Answer

Rescaling features to a comparable range or distribution

Answer

Deleting rows that contain missing values

Answer

Removing all categorical columns from a dataset

Answer

Converting numbers into text labels

Question 2

Write the formula for Min-Max normalization that scales a value x to the range [0, 1].

Accepted Answer

x' = (x - min) / (max - min)

Answer

x' = (x - mean) / std

Answer

x' = x / max

Answer

x' = (x - max) / (max - min)

Question 3

Write the formula for standardization (z-score normalization).

Accepted Answer

z = (x - mu) / sigma

Answer

z = (x - min) / (max - min)

Answer

z = x / sigma

Answer

z = (x - mu) * sigma

Question 4

After standardization (z-score), what are the mean and standard deviation of the transformed feature?

Accepted Answer

Mean 0, standard deviation 1

Answer

Mean 1, standard deviation 0

Answer

Mean 0, standard deviation 0

Answer

Min 0, max 1

Question 5

After Min-Max scaling to [0, 1], what are the minimum and maximum values of the transformed feature?

Accepted Answer

Minimum 0 and maximum 1

Answer

Mean 0 and standard deviation 1

Answer

Minimum -1 and maximum 1

Answer

Minimum 1 and maximum 100

Question 6

A feature has values with min = 10 and max = 50. Apply Min-Max scaling to the value x = 30.

Accepted Answer

0.5

Answer

0.3

Answer

0.6

Answer

0.2

Question 7

A feature has mean = 20 and standard deviation = 5. What is the z-score of the value x = 30?

Accepted Answer

2

Answer

0.5

Answer

10

Answer

1.5

Question 8

Which scaling method is generally more sensitive to outliers: Min-Max scaling or standardization?

Accepted Answer

Min-Max scaling

Answer

Standardization

Answer

Both are completely immune to outliers

Answer

Neither is affected by outliers

Question 9

Does standardization change the shape of a feature's distribution (for example, turn a skewed distribution into a normal one)?

Accepted Answer

No, it only shifts and rescales; the shape is unchanged

Answer

Yes, it always makes the data normally distributed

Answer

Yes, it removes all skewness automatically

Answer

No, because it deletes outliers first

Question 10

In scikit-learn, why must you fit the scaler on the training data only and reuse it to transform the test data?

Accepted Answer

To avoid data leakage from the test set into training

Answer

Because the test set has no minimum or maximum

Answer

To make the test set larger

Answer

Because scalers cannot be reused

Question 11

Which scikit-learn class performs z-score standardization, and which performs Min-Max scaling?

Accepted Answer

StandardScaler and MinMaxScaler

Answer

Normalizer and Binarizer

Answer

LabelEncoder and OneHotEncoder

Answer

PCA and StandardScaler

Question 12

Why do distance-based algorithms like K-Nearest Neighbors require feature scaling?

Accepted Answer

Distances are dominated by large-range features without scaling

Answer

KNN cannot handle decimal numbers

Answer

Scaling reduces the number of neighbors needed

Answer

KNN requires all features to be integers

Question 13

How does feature scaling affect the convergence of gradient descent?

Accepted Answer

It speeds convergence by making the loss surface more symmetric

Answer

It slows convergence by adding noise

Answer

It has no effect on gradient descent

Answer

It prevents gradient descent from ever converging

Question 14

Name two algorithms that generally do NOT require feature scaling.

Accepted Answer

Decision trees and random forests

Answer

KNN and SVM

Answer

Logistic regression and neural networks

Answer

K-means and PCA

Question 15

How can you scale a feature to a custom range such as [-1, 1] using Min-Max scaling?

Accepted Answer

Use MinMaxScaler(feature_range=(-1, 1))

Answer

Use StandardScaler() with mean 0

Answer

Divide every value by 2

Answer

Subtract 1 from every raw value

Question 16

What does a negative z-score indicate about a data value?

Accepted Answer

The value is below the mean

Answer

The value is above the mean

Answer

The value equals the maximum

Answer

The value is invalid

Question 17

If you Min-Max scale data to [0, 1] and then encounter a new test value larger than the training maximum, what happens?

Accepted Answer

The scaled value exceeds 1

Answer

The scaler raises an error and stops

Answer

The value is automatically set to exactly 1

Answer

The value becomes negative

Question 18

Is Min-Max scaling a linear transformation? Briefly justify your answer.

Accepted Answer

Yes, it has the form x' = a*x + b with constant a and b

Answer

No, it squares the input

Answer

No, it takes a logarithm

Answer

Yes, but only for negative values

Question 19

What is the difference between standardization and scikit-learn's Normalizer?

Accepted Answer

StandardScaler scales columns; Normalizer scales each row to unit norm

Answer

They are identical operations

Answer

Normalizer scales columns; StandardScaler scales rows

Answer

Both scale only the target variable

Question 20

When should you generally prefer Min-Max scaling over standardization?

Accepted Answer

When you need a fixed bounded range and have few outliers

Answer

When the data has many extreme outliers

Answer

When you want mean exactly 0 and std exactly 1

Answer

Only for tree-based models

Question 21

What problem occurs in Min-Max scaling if a feature has only one unique value (all values equal)?

Accepted Answer

Division by zero because max minus min equals 0

Answer

The feature is automatically duplicated

Answer

The mean becomes infinite

Answer

Nothing, all values become 1

Question 22

In a machine learning pipeline, should scaling be done before or after splitting into train and test sets?

Accepted Answer

After splitting, fitting only on the training set

Answer

Before splitting, on the whole dataset

Answer

It does not matter at all

Answer

Only on the test set

Question 23

Given the data [2, 4, 6, 8], what is the Min-Max scaled value of 4?

Accepted Answer

About 0.33

Answer

0.5

Answer

0.25

Answer

0.4

Question 24

Why is standardization sometimes called 'mean centering and unit variance' scaling?

Accepted Answer

It centers data at mean 0 and rescales to variance 1

Answer

It centers data at the median and unit range

Answer

It only divides by the maximum

Answer

It removes the mean and the outliers

Question 25

Should you scale the target variable (y) in a regression problem?

Accepted Answer

Usually only X is scaled; scaling y is optional and needs inverse-transform

Answer

You must always scale y the same as X

Answer

y can never be scaled under any circumstance

Answer

Scaling y removes the need to scale X

Question 26

After standardizing a feature, roughly what fraction of values fall between z = -1 and z = +1 if the data is approximately normal?

Accepted Answer

About 68%

Answer

About 50%

Answer

About 95%

Answer

About 99.7%

Question 27

Does Min-Max scaling preserve the relative order (ranking) of data values?

Accepted Answer

Yes, it preserves the order of values

Answer

No, it reverses the order

Answer

No, it randomly reorders values

Answer

Only for positive values

Question 28

How would you manually standardize a NumPy array x to z-scores?

Accepted Answer

z = (x - x.mean()) / x.std()

Answer

z = x / x.max()

Answer

z = (x - x.min()) / (x.max() - x.min())

Answer

z = x.mean() - x

Question 29

Why might Min-Max scaling be a poor choice when your data contains extreme outliers?

Accepted Answer

An outlier compresses normal values into a tiny interval

Answer

Outliers are automatically removed by Min-Max

Answer

Min-Max multiplies outliers by zero

Answer

Outliers make all values equal to 1

Question 30

What is RobustScaler and when is it useful?

Accepted Answer

A scaler using median and IQR, good for outlier-heavy data

Answer

A scaler that deletes all outliers

Answer

A scaler that only works on text

Answer

A scaler identical to MinMaxScaler

Test yourself