Clustering

Question 1

What topics are under dissimilarities?

Answer 1

Continuous dissims.
Binary dissims.
3 Rules
Dissim matricies

Question 2

What are two types of continuous dissims?

Answer 2

Euclidean

Manhattan

Question 3

What are two types of binary dissims?

Answer 3

Simple

Jaccard

Question 4

What is the Simple binary dissimilarity equation?

Answer 4

d(x,y)=(a+d)/(a+b+c+d)

Question 5

What is the Jaccard binary dissimilarity equation?

Answer 5

d(x,y)=a/(a+b+c)

Question 6

What are the three rules for dissimilarities?

Answer 6

d(x,y)>=0, if x=y then d=0
d(x,y)=d(y,x)
d(x,y)>=d(x,z)+d(x,y)

Question 7

What are the key features of dissimilarity matricies?

Answer 7

0’s along the diagonal
Usually factors are standardized
Usually factors are scaled
Symmetric

Question 8

What topics are important to hierarchical clustering?

Answer 8

5 steps
Linkage
Chaining
Number of groups

Question 9

What are the five steps in hierarchical clustering?

Answer 9

1. Each obs. into its own group
2. Pair nearest obs.
3. One less group now!
4. Pair next nearest obs.
5. Repeat until no individual obs left

Question 10

What types of linkage are there?

Answer 10

Simple
Complete
Average

Question 11

What is the formula for simple linkage?

Answer 11

d(A,B)=min x€A,y€B,d(x,y)

Question 12

What is the formula for complete linkage?

Answer 12

d(A,B)=max x€A,y€B,d(x,y)

Question 13

What is important to know about the number of groups in hierarchical clustering?

Answer 13

The dendrogram should be cut along the y-axis, this divides the total obs. into groups below the line.

Question 14

What is chaining?

Answer 14

Where observations are individually appended to the same group one after another -results in poor model.

Question 15

What topics are under partitioning methods?

Answer 15

6 steps
K-means
Starting position
Number of groups

Question 16

What are the 6 steps to partitioning methods?

Answer 16

1. Data is partitioned into groups
2. Centroids of groups are calculated
3. Each obs. distance from centroid is calculated
4. Move obs to new group if necessary
5. If no moves are made, stop, local min found
6. Repeat from 2 otherwise

Question 17

What is important to knows about k means?

Answer 17

It is iterative
It is computationally efficient (+)
It results in local minimums (-)

Question 18

What is important to remember about starting positions?

Answer 18

Random point
Random area
Or result from hierarchical cluster

Question 19

How is the number of clusters in the data determined?

Answer 19

Trialling range of clusters (e.g., 1-10)
Plotting sum of squares
Looking for elbow in graph

Question 20

What three topics are under cluster validation?

Answer 20

5 steps
Rand index
Issues with clustering

Question 21

What are the issues with clustering?

Answer 21

Sensitive to the start point
Only works for continuous data
Clusters are all spherical

Question 22

What things are important to note about the rand index?

Answer 22

How to calculate: rand(S1,S2)=A/A+D

Adjusted rand accounts for natural randomness

Question 23

What are the 5 steps in cluster validation?

Answer 23

1. Split data into train/test
2. Run clustering on training data
3. Group test data based on cluster: S1
4. Cluster test individually: S2
5. Cross tabulate S1 and S2