Lecture 16 - Clustering

Question 1

What is clustering?

Answer 1

Given a set of unlabelled training examples, finding a way to partition the examples into classes/groups

Therefore be able to determine the class of any new sample

Question 2

Clustering is also known as ____________ learning

Answer 2

unsupervised

Question 3

What criteria make a good partition for clustering?

Answer 3

Maximise similarity within classes

Minimise similarity between classes

Minimise number of classes created

Maximises ability to predict unknown attribute values from class membership

Question 4

Agglomerative Hierarchy and K-Means method are both methods of?

Answer 4

Clustering

Question 5

What is the basic procedure of agglomerative hierarchical clustering?

Answer 5

Assign each sample to its own cluster

While there are at least X clusters

Find the most similar pair of clusters

Merge them into a new larger cluster

Question 6

What are the prerequisites for agglomerative hierarchical clustering?

Answer 6

Similarity metric for samples

All examples must be available at the start

human analyst to determine optimal number of clusters

Question 7

What is the basic procedure of k-means method?

Answer 7

k = number of clusters to form

Choose k items randomly to be cluster centers

repeat until no item changes clusters {

assign each item to its nearest cluster

set each cluster center to be the mean value of each item in the cluster

}