Chapter 13 Quiz

Question 1

approach which combines multiple supervised models into a supermodel

Answer 1

ensemble

Question 2

what will averaging multiple models yield?

Answer 2

a more precise answer than found individually

Question 3

four ways to make ensembles

Answer 3

simple averaging or voting
combining predictions (median)
combining classifications
combining propensities

Question 4

generate multiple random samples, run algorithm on each same and produce scores

Answer 4

bagging

Question 5

fit model to data, draw sample with misclassified records that have a higher probability of selection, fit model to new sample

Answer 5

boosting

Question 6

generating rank-ordered list of candidate models with automatic parameter tuning

Answer 6

automated machine learning

Question 7

what does creating an ensemble do?

Answer 7

reduces the root mean squared error

Question 8

what is the relationship between bias, variance, covariance, and correlation

Answer 8

bias <– variance <— covariance
covariance affects correlation

Question 9

advantages of ensembles

Answer 9

more precise (especially for low/negative correlations)

Question 10

weaknesses of ensembles

Answer 10

uncorrelated samples result in uncorrelated predictions
computational resources
blackbox model

Question 11

advantages of AutoML

Answer 11

good starting point that can be tweaked to speed up analysis

Question 12

weaknesses of AutoML

Answer 12

do not understand business problem/dataset