Clustering

Question 1

Hard clustering

Answer 1

Every object is a member of one one cluster.

Question 2

Soft clustering

Answer 2

Objects can be members in more than one cluster.

Question 3

Hierarchical Clustering

Answer 3

Pairs of most-similar clusters are iteratively linked until all objects are in a clustering relationship.

Question 4

Non-hierachical clustering

Answer 4

Results in flat clusters of “similar” documents

Question 5

Proximity Measure

Answer 5

Gives a measure of proximity between two documents. As measures are symmetric, the matrix is a triangle.

Question 6

Raw Overlap

Answer 6

|X intersection Y|

Question 7

Dice’s coefficient

Answer 7

2|X intersection Y|/(|X| + |Y|)

Question 8

Jaccard’s Coefficient

Answer 8

|X intersection Y|/|X union Y|

Question 9

Overlap coefficient

Answer 9

|X intersection Y|/minimum(|X|,|Y|)

Question 10

Cosine overlap

Answer 10

|X intersection Y|/(sqrt(|X|) * sqrt(|Y|))

Question 11

Single Link Function

Answer 11

Similarity of two most similar members.

Question 12

Complete Link Function

Answer 12

Similarity of two least similar members.

Question 13

Group Average Function

Answer 13

Avg. similarity of each pair of group members.

Question 14

Single Link vs. Complete Link

Answer 14

O(N^2) vs. O(N^3)

Question 15

Centroid

Answer 15

Average vector of cluster

Question 16

Medoid

Answer 16

Representative vector closes to centroid

Question 17

Measure of cluster quality

Answer 17

Mean square distance from each data point to its nearest centre should be minimal.

Question 18

Other Algorithms

Answer 18

Too complex and mathematical to put in flashcard - look up in notes pages 118-123.