Machine Learning

Question 1

Supervised learning

Answer 1

labeled training data to guide the ML program toward superior forecasting accuracy

Question 2

Unsupervised learning

Answer 2

the ML program is not given labeled training data

Question 3

Deep learning

Answer 3

used for complex tasks such as image recognition, natural language processing, and so on

Question 4

reinforced learning

Answer 4

Programs that learn from their own prediction errors

Question 5

neural networks

Answer 5

a group of ML algorithms applied to problems with significant nonlinearities

Question 6

Supervised Learning Types

Answer 6

Regression (Continuous)
Classification (Categorical)
Neural Networks
Deep Learning
Reinforcement LEarning

Question 7

Unsupervised Learning Types

Answer 7

Dimensionality Reduction
Clustering
Neural Networks
Deep Learning
Reinforcement LEarning

Question 8

Overfitting

Answer 8

when a large number of features are included in the data sample

Question 9

cross validation

Answer 9

estimates out-of-sample error rates directly from the validation sample.

Question 10

To measure how well a model generalizes, data analysts create three nonoverlapping data sets

Answer 10

(1) training sample
(2) validation sample
(3) test sample

Question 11

Data scientists then decompose these errors into the following:

Answer 11

• Bias error.
• Variance error.
• Base error.

Question 12

Bias error.

Answer 12

This is the in-sample error resulting from models with a poor fit.

Question 13

Variance error.

Answer 13

This is the out-of-sample error resulting from overfitted models that do not generalize well.

Question 14

Base error.

Answer 14

These are residual errors due to random noise.

Question 15

k-fold cross validation

Answer 15

the sample is randomly divided equally into k parts. The training sample comprises (k − 1) parts, with one part left for validation. Error is then measured for the model in each of the parts.

Question 16

common supervised ML algorithms

Answer 16

Penalized regressions.

Support vector machine (SVM).

K-nearest neighbor (KNN).

Classification and regression trees (CART).

Ensemble and Random Forest.

Question 17

Penalized regressions

Answer 17

reduce the problem of overfitting by imposing a penalty based on the number of features used by the model.

Question 18

Least absolute shrinkage and selection operator (LASSO).

Answer 18

popular penalized regression model. In addition to minimizing SSE, LASSO minimizes the sum of the absolute values of the slope coefficients.

Question 19

Support vector machine (SVM)

Answer 19

linear classification algorithm that separates the data into one of two possible classifiers (e.g., sell vs. buy).

Question 20

K-nearest neighbor (KNN)

Answer 20

used to classify an observation based on nearness to the observations in the training sample

Question 21

Classification and regression trees (CART)

Answer 21

Classification trees assign observations to one of two possible classifications at each node.

Question 22

Ensemble and Random Forest.

Answer 22

Ensemble learning is the technique of combining predictions from multiple models rather than a single model.

Random forest is a variant of classification trees whereby a large number of classification trees are trained using data bagged from the same data set.

Question 23

common unsupervised ML algorithms

Answer 23

Principal component analysis (PCA).

Clustering.

Question 24

Principal component analysis (PCA)

Answer 24

summarizes the information in a large number of correlated factors into a much smaller set of uncorrelated factors.

Question 25

Clustering.

Answer 25

clustering is the process of grouping observations into categories based on similarities in their attributes (called cohesion)

Question 26

K-means clustering

Answer 26

partitions observations into k nonoverlapping clusters, where k is a hyperparameter (i.e., set by the researcher).

Question 27

Hierarchical clustering

Answer 27

builds a hierarchy of clusters without any predefined number of clusters

Question 28

agglomerative (or bottom-up) clustering

Answer 28

start with one observation as its own cluster and add other similar observations to that group, or form another nonoverlapping cluster.

Question 29

divisive (or top-down) clustering

Answer 29

starts with one giant cluster, and then it partitions that cluster into smaller and smaller clusters.

Question 30

Neural Networks

Answer 30

constructed as nodes connected by links

Question 31

Deep learning networks (DLNs)

Answer 31

neural networks with many hidden layers

Question 32

Reinforcement learning (RL)

Answer 32

have an agent that seeks to maximize a defined reward given defined constraints