3: Machine Learning

Question 1

Data are shuffled randomly and then divided into k equal subsamples.
One sample is saved to be used as validation sample, and the other k-1 samples are used as training samples

Answer 1

K-fold cross validation

Question 2

Technique of combining predictions from a number of models, with the objective of canceling out noise

Answer 2

Ensemble Learning

Results in: more accuracy & stable predictions (vs single model)

Question 3

• Nodes connected by links
• Useful in: Supervised Regression & Classification models
• Works well in presence of: nonlinearities & complex interactions among variables
• Recognizes: patterns, clusters, and classifies

Answer 3

Neural Networks

Question 4

Unsupervised Neural Networks with many hidden layers (often >20), and reinforcorced learning learn from their own prediction errors

Used for: complex tasks; image, pattern, & character recognition

Answer 4

Deep Learning Networks

Question 5

• Algorithm learns from success & mistakes
• Seeking to maximize reward and minimize punishment
• Defined constraints

Answer 5

Reinforcement Learning

Question 6

Inputs & outputs are identified for the computer, and the algorithm uses this labeled training data to model relationships

Answer 6

Supervised Learning

Question 7

Computer is provided unlabeled data that the algorithm uses to determine the structure of the data

Answer 7

Unsupervised Data

Question 8

Least Absolute Shrinkage and Selection Operator (LASSO) is useful in building:

Penalized regression model

Answer 8

Parsimonious models, through feature reduction

Question 9

K-Nearest Neighbor, investment application includes:

Used in: classification & regression

Answer 9

• predicting bankrupcty
• assigning bond ratings class
• predicting stock prices
• creating customized indicies

Question 10

Random Forest investment applications include:

Answer 10

• factor based asset allocation
• prediction models for IPO success

Question 11

Linear relationships

A penalized regression model tries to use a limited number of most important features that…

Answer 11

explain the variation in the dependent variable

Example: monthly returns on 100 stocks

Question 12

Overfitting occurs when:
Bias error:
Variance error:

Answer 12

when model fits the training too well
Bias error: low
Variance error: high

displaying non linear characteristics

Question 13

Generalize is the degree to which the model retains it’s explanatory power when:

Answer 13

predicting out of sample

Question 14

Bias error is the degree to which:

Answer 14

the model fits the training data

Question 15

Variance error shows how much the model responds to:

Answer 15

new data

Question 16

How to prevent overfitting:

Answer 16

• don’t let model become too complex
• proper data sampling using cross validation (k-fold)

Question 17

Complexity Reduction:

Answer 17

Dimensional Reduction
Use: PCA

Question 18

With supervised data, the training data contains:

Answer 18

ground truth

Question 19

Supervised ML algorithm

Classification focuses on sorting observation into:

Answer 19

distinct categories:
* pass or failure

Question 20

Regression based uses:

Answer 20

continuous variables

Question 21

Regression:

CART & Random forests are used for:

Answer 21

complex & non-linear

Question 22

Classified unsupervised data:

K-means is used for:

Answer 22

complex & linear data
with a known number of k clusters