1. Given the following silhouette scores, how many clusters should you choose?

Silhouette scores of 2 clusters:

1. 0.8

2. 0.6

Silhouette scores of 3 clusters:

1. 0.6

2. 0.5

3. 0.4

Silhouette scores of 4 clusters:

1. 0.7

2. 0.8

3. 0.6

4. 0.7

Given the following silhouette scores, how many clusters should you choose?

Silhouette scores of 2 clusters:

1. 0.8

2. 0.6

Silhouette scores of 3 clusters:

1. 0.6

2. 0.5

3. 0.4

Silhouette scores of 4 clusters:

1. 0.7

2. 0.8

3. 0.6

4. 0.7

1 or 2

1

1 or 3

3

2

1. or 3

1. What risks could bootstrapping result in, with regard to ensemble learning? Select the best answer. What risks could bootstrapping result in, with regard to ensemble learning? Select the best answer.

Inefficiency due to looping through observations

Excluding some observations altogether

Impure splits

Computational complexity

1. In which of the following scenarios would ward linkage work best? Select all that apply

Noisy data

Cylindrical data

Cleanly separable data

Highly random data

1. In which of these scenarios should we use a decision tree over another model?

Predicting which products a customer is likely to buy

Blocking a credit card based on fraudulent transactions

Predicting which ads to display to a customer

1. A decision tree with high depth is likely to suffer from ________. Select the best answer.

Underfitting

High RMSE

Overfitting

Low accuracy

1. In which of the following scenarios would you expect DBSCAN to work better than other clustering techniques?

When trying to cluster time series data

When trying to cluster customer data where a lot of features are available

When trying to cluster categorical data

When trying to cluster customer data where a lot of features are not available

1. In model-based recommendation engines, the algorithm tests iterative values of feature matrices until it's able to predict a ratings matrix accurately. What happens when the SSE is high?

The algorithm attempts one final iteration

The algorithm reverts to the last set of values

The algorithm tries a new set of feature values

Nothing - a high SSE means the ratings are fairly accurate

1. Which of the following is a scenario where the kernel trick in a support vector model is helpful? Select the best answer.

When data follows a straight line pattern

When data has clearly defined, round clusters

When data is scattered in an undefined pattern

When data is not linearly separable

1. How does shrinkage help combat overfitting in boosting algorithms?

It removes unnecessary branches from tree models

It reduces the influence of each model on the overall result

It regularizes insignificant coefficients to 0

It minimizes insignificant coefficients

1. What are support vectors? Select the best answer.

The examples furthest from the hyperplane.

The examples which influence the position of the hyperplane.

The examples which influence the decision boundary.

The examples closest to the hyperplane.

1. How does an AdaBoost model make a prediction?

Soft voting for classification and averaging for regression problems

Majority voting for classification and regression problems

Majority voting for classification and averaging for regression problems

Weighted average voting for classification and regression problems

1. What depth of trees does an AdaBoost model use?

1

2

Any depth

0