Unit 6 - Applications of clustering

Question 1

What are the five steps of clustering analysis?

Answer 1

1) Stating the business problem
2) Methodological design
3) Cluster generation (preliminary analysis)
4) Cluster interpretation and evaluation
5) cluster validation and profiling

Question 2

What are some of the methodological design?

Answer 2

• How to collect data?
• Which clustering technique to use? (hierarchical, partioning, SOM or twoStep)
• Which proximity measure to use?
• Should data be standardized?

Question 3

What is cluster generation (preliminary analysis)?

Answer 3

• Explore data using summary statistics/visualization
• Inclusion or exclusion of outliers must be considered
  carefully
• Compare different solutions

Question 4

What is cluster interpretation and evaluation?

Answer 4

• Performs a more detailed examination of the clustering solution
• May use statistical tests to compare clusters’ centroids to find distinct clusters.

Question 5

What is cluster validation and profiling?

Answer 5

• Clustering techniques may generate clustering solutions even when no natural groupings exist
• Important to assess cluster validity (Two ways)
  > Use other clustering methods to double-check
  > Use variables not included in the analysis
  (e.g., Customer groups related to income categories,
  class label given in the GBA question)

Question 6

What is market segmentation?

Answer 6

• Divide a market into different segments
• Help to better understand consumers’ profiles & needs
• Each segment respond similarly to a marketing campaign
• Allows suppliers to differentiate their goods and services
• Lead to improved marketing efforts and effectiveness.

Question 7

What are some variables used by market segmentation?

Answer 7

• Geographical—country, region, climate, area size,…
• Demographic—age, nationality, ethnicity, gender,…
• Health status—primary diagnosis, stage of illness,…
• Psychological—personality, value, taste, lifestyle,…
• Socio-cultural—cultures, religion, social class,…
• Behavioural—brand loyalty, personal habits, …
• Product knowledge—level (non-user,current user, expert)