quiz 4

Question 1

What is Predictive Data Mining?

Answer 1

A form of supervised machine learning that uses input data to predict a known outcome.

Question 2

What are the steps in the CRISP-DM process?

Answer 2

• Business Understanding
• Data Understanding
• Data Sampling
• Data Preparation
• Data Partitioning
• Model Construction
• Model Evaluation
• Deployment

Question 3

What does a Confusion Matrix represent?

Answer 3

It tracks how well the model is classifying outcomes.

Question 4

What are the components of a Confusion Matrix?

Answer 4

• True Positive (TP)
• False Negative (FN)
• False Positive (FP)
• True Negative (TN)

Question 5

What is the formula for Accuracy?

Answer 5

Accuracy = (TP + TN) / Total

Question 6

What does Precision measure?

Answer 6

Precision = TP / (TP + FP)

Question 7

What does Recall / Sensitivity measure?

Answer 7

Recall = TP / (TP + FN)

Question 8

What is Specificity?

Answer 8

Specificity = TN / (TN + FP)

Question 9

What is the F1 Score?

Answer 9

The harmonic mean of Precision and Recall.

Question 10

What is the default cutoff value in classification?

Answer 10

0.5

Question 11

What happens when the threshold is lowered?

Answer 11

More ‘yes’ predictions, higher recall, lower precision.

Question 12

What is a Lift Chart?

Answer 12

A visual tool that compares a model’s ability to rank positives vs. a random guess.

Question 13

What does ROC Curve plot?

Answer 13

True Positive Rate (Recall) vs False Positive Rate (1 – Specificity).

Question 14

Why not use linear regression for classification?

Answer 14

Linear regression produces continuous outputs and doesn’t bound predictions between 0 and 1.

Question 15

What transformation does logistic regression use?

Answer 15

The logit transformation.

Question 16

What is the logistic model equation?

Answer 16

ln(p / (1 - p)) = β0 + β1x1 + β2x2 + …

Question 17

What is k-Nearest Neighbours (k-NN)?

Answer 17

A method that classifies new observations based on the closest training points.

Question 18

What distance metric is commonly used in k-NN?

Answer 18

Euclidean distance.

Question 19

What is the effect of choosing k = 1 in k-NN?

Answer 19

Highly sensitive, may overfit.

Question 20

What are key features of Classification and Regression Trees (CART)?

Answer 20

• Non-parametric model
• Tree-based structure
• Splits based on informative features

Question 21

What is overfitting in the context of CART?

Answer 21

Fitting the training data too well, resulting in high variance.

Question 22

What is the purpose of pruning in CART?

Answer 22

To improve generalisability based on validation error.

Question 23

What is an Ensemble Method?

Answer 23

Combines multiple weak models to create a stronger model.

Question 24

What is Monte Carlo Simulation?

Answer 24

A technique to model uncertainty using random values from probability distributions.

Question 25

What does the output of a Monte Carlo Simulation represent?

Answer 25

A range of possible outcomes.

Question 26

What function is used to generate uniform random numbers in Excel?

Answer 26

RAND()

Question 27

What is the Excel function to simulate a normal distribution?

Answer 27

NORM.INV(RAND(), μ, σ)

Question 28

What are the key variables in the Sanotronics case?

Answer 28

* Labour cost * Parts cost * Demand

Question 29

What type of distribution is used for Parts cost in the Sanotronics case?

Answer 29

Uniform distribution.

Question 30

What is the goal of the Land Shark case?

Answer 30

Simulate whether a bid will win a property auction.

Question 31

What are the inputs for the Land Shark case simulation?

Answer 31

* Estimated property value * Random number of competing bidders * Competitor bid values

Question 32

What is a limitation of simulation?

Answer 32

Results are estimates, not guarantees.