IFN580 Week 8 Clustering

Question 1

What is clustering?

Answer 1

A ML technique that groups similar data points together

Question 2

K-means terminates when?

Answer 2

When the cluster centres for the current iteration are identical to the previous iteration

Question 3

What is hierarchical clustering?

Answer 3

building a hierarchy of clusters represented by a dendogram

Question 4

How are hierarchical clusters built?

Answer 4

by splitting or merging clusters based on distance between them

Question 5

Which clustering algorithm initially assumes that each instance represents a single cluster?

Answer 5

Agglomerative

Question 6

List some of the distance measures that are used in clustering.

Answer 6

Manhattan Distance
Euclidean Measure
Minkowski Distance
Dot Point
Cosine Similarity

Question 7

How can the distance between missing values be computed?

Answer 7

Assume they are maximally distant

Question 8

Unsupervised evaluation can be internal or external. Which of the following is an internal method for evaluating alternative clusterings produced by the K-Means algorithm?

Answer 8

Compare the sum of squared error differences

Question 9

What is the range for silhouette scores?

Answer 9

between -1 and 1

Question 10

What are silhouette scores?

Answer 10

calculates the quality of a clustering solution using intra-cluster and inter-cluster distances

1 = clusters are well apart
0 = clusters are overlapping
-1 = point is in wrong cluster

Question 11

How to determine what is the optimal number of clusters when
using the using the elbow method?

Answer 11

the point at which the reduction starts to level off