Last Minute Information

Question 1

What are the three types of missing values?

Answer 1

1. MAR
2. MNAR
3. MCAR

Question 2

What does MAR stand for?

Answer 2

Missing at Random

Question 3

What does MNAR stand for?

Answer 3

Missing Not At Random

Question 4

What does MCAR stand for?

Answer 4

Missing Completely at Random

Question 5

What is Missing at Random Data?

Answer 5

Data that is more likely to be missing for some data objects.

Question 6

What is Missing Completely at Random Data?

Answer 6

Data that is missing purely by chance, and all data objects have the same chance of it being missing.

Question 7

What is Missing Not At Random Data?

Answer 7

Data that is missing, and it is known which data objects have missing data.

Question 8

What are 4 approaches to dealing with Missing Values?

Answer 8

1. Keep as-is
2. Remove Columns
3. Remove Rows
4. Impute Values

Question 9

What are three types of outliers?

Answer 9

1. Data errors
2. Legitimate Values
3. Fraudulent Entries

Question 10

What are 4 ways to deal with outliers?

Answer 10

1. Do Nothing
2. Replace with upper or lower cap
3. Log Transformation
4. Remove data objects with outliers

Question 11

What are the two types of data errors?

Answer 11

1. Random Errors
2. Systemic Errors

Question 12

What are the two types of data transformation?

Answer 12

1. Normalisation
2. Standardisation

Question 13

What does standardisation do?

Answer 13

Rescale the data to have a mean of 0 and a standard deviation of 1

Question 14

What doe Normalisation do?

Answer 14

Rescale the data to have a common scale, often [0, 1]

Question 15

What are support vectors?

Answer 15

The points in an SVM that lie closest to the boundary.

Question 16

What does DBSCAN stand for?

Answer 16

Density-based Spatial Clustering of Applications with Noise

Question 17

What does PCA stand for?

Answer 17

Principle Component Analysis

Question 18

How does DBSCAN cluster?

Answer 18

It finds areas of high density and expands the clusters from there.

Question 19

What does t-SNE stand for?

Answer 19

t-distributed stochastic neighbour embedding

Question 20

What does UMAP stand for?

Answer 20

Uniform Manifold Approximation and Projection

Question 21

What are 4 ways to measure cluster difference?

Answer 21

1. Simple Linkage
2. Complete Linkage
3. Average Linkage
4. Centroid Linkage

Question 22

What is Ward’s Method?

Answer 22

Its a method that decides whether clusters should be joined based on if it reduces the total distance from centroids.

Question 23

How does DBSCAN create clusters?

Answer 23

For any point in any cluster, the point density around that point has to exceed a given threshold.

Question 24

Which two hyperparameters define density in DBSCAN?

Answer 24

1. eps (epsilon)
2. minPoints

Question 25

What is a core point?

Answer 25

A point that has more than the specified number of points within eps. These points are at the interior of a cluster.

Question 26

What is a Border point?

Answer 26

A point that has fewer than minpoints within eps but is within a core points neighbourhood.

Question 27

What is a noise point?

Answer 27

A point that is not a border point or a core point.

Question 28

What are advantages of DBSCAN?

Answer 28

1. Resistant to Noise 2. Can handle clusters of different shapes and sizes

Question 29

How does a Gaussian Mixture Model represent a cluster?

Answer 29

A cluster is defined by a probability distribution over it.