Predicting numbers (regression) quiz

Question 1

What is 'regression' in machine learning?

Accepted Answer

Predicting a continuous number from input data

Answer

Sorting data points into labeled categories

Answer

Finding and removing duplicate records in a table

Answer

Compressing images to save storage space

Question 2

What is the 'line of best fit' in regression?

Accepted Answer

A straight line that best represents the overall trend in the data

Answer

A line connecting the very first and very last data points

Answer

A line that passes exactly through every single data point

Answer

The shortest horizontal line that can be drawn on the graph

Question 3

What are 'features' and 'labels' in a regression problem?

Accepted Answer

Features are input values used to predict; labels are the values being predicted

Answer

Features are the answers; labels are the questions

Answer

Both features and labels mean exactly the same thing in machine learning

Answer

Labels are simply the column names in a dataset

Question 4

What does it mean to 'train' a regression model?

Accepted Answer

Showing the model many input-output pairs so it learns the underlying pattern

Answer

Writing the Python code that defines the model structure

Answer

Evaluating the model's performance on brand-new unseen data

Answer

Deleting incorrect or missing rows from the raw dataset

Question 5

Which of the following is a real-world example of a regression problem?

Accepted Answer

Predicting tomorrow's temperature in degrees Celsius

Answer

Deciding whether an email is spam or not spam

Answer

Sorting photos into 'cat' and 'dog' folders

Answer

Detecting whether a bank transaction is fraudulent

Question 6

What does the y-intercept 'b' represent in the regression equation y = mx + b?

Accepted Answer

The predicted value of y when the input feature x equals zero

Answer

The rate at which y increases for every one-unit increase in x

Answer

The average of all y values across the entire training dataset

Answer

The exact point where the regression line crosses the horizontal x-axis

Question 7

What does a positive correlation between two variables indicate?

Accepted Answer

As one variable increases, the other variable tends to increase as well

Answer

As one variable increases, the other variable tends to decrease

Answer

The two variables are completely unrelated to each other

Answer

Both variables remain constant regardless of the other

Question 8

What does it mean when a regression model has a negative slope?

Accepted Answer

As the input feature increases in value, the model's predicted output decreases

Answer

The model is systematically making all its errors in the wrong direction

Answer

The y-intercept must also be a negative value

Answer

The model is too simple and is underfitting the data

Question 9

In the linear regression equation y = mx + b, what does the slope 'm' represent?

Accepted Answer

How much the prediction changes for every one-unit increase in the input feature

Answer

The predicted value of y when the input feature x equals zero

Answer

The speed at which the computer trains the model

Answer

The average of all y values in the training dataset

Question 10

What are 'residuals' in regression, and why do they matter?

Accepted Answer

The difference between the actual value and the model's predicted value

Answer

The unused columns that were dropped during data cleaning

Answer

The features that were removed because they were irrelevant

Answer

The random numbers used to initialise the model's parameters

Question 11

What is Mean Absolute Error (MAE) and how is it calculated?

Accepted Answer

The average of the absolute differences between predictions and actual values

Answer

The single largest error the model made on any one data point

Answer

The average of the squared differences between predictions and actual values

Answer

The percentage of data points the model predicted correctly

Question 12

Why do we split data into a 'training set' and a 'test set' before building a regression model?

Accepted Answer

To fairly evaluate whether the model performs well on genuinely unseen data

Answer

To reduce the file size of the dataset before processing

Answer

To make the model training process run faster

Answer

To automatically remove duplicate rows from the dataset

Question 13

What is 'overfitting' in a regression model?

Accepted Answer

The model performs very well on training data but poorly on new, unseen data

Answer

The model is too simple and cannot capture any pattern in the data

Answer

The dataset has too many rows for the computer to process efficiently

Answer

The model takes too many hours to finish training

Question 14

What is 'underfitting' in a regression model?

Accepted Answer

The model is too simple and misses the true pattern in the data

Answer

The model has memorised the training data so perfectly that it cannot generalise

Answer

The model is so complex that it requires too much memory to run

Answer

The test set is too small relative to the training set

Question 15

What is the difference between interpolation and extrapolation when making regression predictions?

Accepted Answer

Interpolation means predicting within the training data range; extrapolation means predicting outside it

Answer

Interpolation means predicting outside the data range; extrapolation means predicting within it

Answer

Both terms simply mean making predictions on the test set

Answer

Extrapolation is always more accurate because it tests the model more rigorously

Question 16

Why might adding more relevant features improve a regression model's accuracy?

Accepted Answer

Relevant features give the model more useful information, helping it predict more accurately

Answer

More features always make the model worse and harder to interpret

Answer

More features make the dataset larger, automatically speeding up training

Answer

Adding enough features removes the need to evaluate the model on a test set

Question 17

What is Mean Squared Error (MSE) and how does it differ from Mean Absolute Error (MAE)?

Accepted Answer

MSE squares residuals before averaging, penalising large errors much more heavily than MAE does

Answer

MSE takes the absolute value of errors while MAE squares them before averaging

Answer

MSE and MAE always produce identical numerical values for the same predictions

Answer

MAE is used only for classification problems, not regression

Question 18

Why is feature scaling important when training some regression models?

Accepted Answer

It puts all features on a similar numerical range, preventing large-valued features from dominating

Answer

It removes duplicate rows from the dataset before training begins

Answer

It automatically converts categorical text features into numbers

Answer

It increases the total number of features available to the model

Question 19

How can outliers negatively affect a regression model?

Accepted Answer

Outliers can pull the regression line away from the true trend in the data, reducing accuracy

Answer

Outliers have no effect because all modern regression algorithms automatically ignore extreme values

Answer

Outliers improve the model by adding useful diversity to the training data

Answer

Outliers are automatically detected and removed by all standard regression algorithms

Question 20

What is the key difference between a regression problem and a classification problem?

Accepted Answer

Regression predicts continuous numbers while classification predicts discrete categories

Answer

Regression uses neural networks while classification always uses decision trees

Answer

Classification always requires more training data than regression

Answer

Regression can only use one input feature while classification can use many

Question 21

What ethical issue can arise when using regression to make important decisions about people?

Accepted Answer

A model trained on historically biased data will learn and reproduce those biases, causing unfair predictions

Answer

Regression models are always fair because they rely purely on objective mathematics

Answer

Ethical concerns only apply to image recognition models, not numerical regression

Answer

Using a larger dataset automatically eliminates all bias from a regression model

Question 22

In scikit-learn, how do you train a simple linear regression model and use it to make predictions?

Accepted Answer

from sklearn.linear_model import LinearRegression

Answer

from sklearn.regression import LinearModel

Answer

from sklearn.models import Regression

Answer

from sklearn.fit import LinearRegression

Question 23

What is 'gradient descent' in simple terms?

Accepted Answer

A step-by-step method that repeatedly adjusts model parameters to minimise prediction error

Answer

A method for automatically splitting data into training and test sets

Answer

A technique for identifying and removing outliers from training data

Answer

A visualisation tool for drawing the regression line on a scatter plot

Question 24

What is a 'loss function' in machine learning, and why is it important during training?

Accepted Answer

A formula that measures how wrong the model's predictions are, which training tries to minimise

Answer

A function that divides data into smaller batches for more efficient processing

Answer

The set of rules that determines which features should be included in the model

Answer

The code that generates random starting values for the model's parameters

Question 25

What does R-squared (R²) measure in a regression model?

Accepted Answer

The proportion of variation in the label that is explained by the model's input features

Answer

The number of features in the model divided by the total number of data points

Answer

The average prediction error of the model measured in original units

Answer

The square root of the Mean Absolute Error

Question 26

What is the 'bias-variance tradeoff' in machine learning?

Accepted Answer

The balance between a model being too simple (underfitting) and too sensitive to training data (overfitting)

Answer

The tradeoff between dataset size and the speed of training the model

Answer

The choice between using MAE or MSE as the model's loss function

Answer

The tradeoff between prediction accuracy and ethical fairness

Question 27

What is polynomial regression, and when is it more useful than simple linear regression?

Accepted Answer

A method that fits a curved line by using powers of the input feature such as x² or x³

Answer

A regression model that only works correctly with datasets containing whole numbers

Answer

A regression model that uses multiple different input features simultaneously

Answer

A special way to predict categorical outputs using a regression framework

Question 28

What is cross-validation, and why is it more reliable than a single train-test split?

Accepted Answer

A technique that trains and evaluates the model multiple times on different data splits and averages results

Answer

A method that uses the entire dataset for training and performs no evaluation at all

Answer

A visual tool for drawing and comparing multiple regression lines on one graph

Answer

An automated method that selects the best subset of features for the model

Question 29

What is 'data leakage', and why is it a serious problem when building regression models?

Accepted Answer

When information from the test set or future data leaks into training, inflating performance scores

Answer

When the training dataset accidentally contains too many duplicate rows

Answer

When a model is trained on data from a different country than where it will be used

Answer

When a dataset is accidentally stored in an insecure cloud storage bucket

Question 30

Why should we never use the test set to choose a model's settings (hyperparameters)?

Accepted Answer

Repeatedly checking the test set effectively trains on it, giving falsely optimistic performance estimates

Answer

Using the test set for tuning significantly speeds up the training process

Answer

The test set is always too small to use reliably for hyperparameter selection

Answer

Hyperparameters are mathematically fixed before any data is collected, so this cannot arise

Test yourself