E6

Question 1

What is clustering?

Answer 1

• Finding groups in data.
• Organizing data into groups such that there is:
(1) high similarity within each group,
(2) low similarity across the groups.

Question 2

Is clustering the same as classification?

Answer 2

No.

• Class labels can be found directly in the data. E.g., blood type.
• Different goals: to “understand” the data better (explore), to organize the information we have.

Question 3

Distance measures

Answer 3

• Euclidean distance
-> physical distance between two data points

• Manhattan distance
-> taxicab distance -> absolute difference

• Jaccard distance
-> treat two objects as sets of characteristics (text mining same word)

• Cosine distance
-> cosine of angle between two vectors (often text mining/recommend)

• Edit distance
-> Levenshtein metric -> autocorrect (spelling mistakes)

Question 4

K-means clustering - how to?

Answer 4

1. Select proximity measure and specify the number of clusters (k).
2. Initiate the process by selecting centroids.
3. Assign the data points to the “nearest” centroid to form a cluster.
4. Calculate the new centroid.
5. Iterate over steps 3 and 4 until the stopping criteria are fulfilled.

Question 5

Strengths k-means

Answer 5

• Simple

- Efficient

Question 6

Weaknesses k-means

Answer 6

• the value of k – how to determine it?
• converges to locally optimal solution
• globular/spherical clusters

Question 7

Hierarchical clustering

Answer 7

Creates a collection of ways to group the points

Question 8

Output of hierarchical clustering

Answer 8

Dendograms

Question 9

Strengths hierarchical clustering

Answer 9

• Clusters can be of any size and shape.

* Does not require to prespecify the number of clusters.

Question 10

Weaknesses hierarchical clustering

Answer 10

• Still need to decide where to split.

* Computationally inefficient.