Online data science course

1
Cluster Analysis:
A practical example

2
Content
Introduction: the necessity to reduce the
complexity
Recall: what cluster analysis does
An example : cluster analysis in consumer
research on fair trade coffee
Discussion

(…)
“Where is the life, we have lost in living?
Where is the wisdom, we have lost in knowledge?
Where is the knowledge, we have lost in information?”
(…)
T. S. Elliot,
Choruses from the Rock
(1888 – 1965)
Intro

(…)
Where is the wisdom, we have lost in knowledge?
Where is the knowledge, we have lost in information?
(…)
“Where is the information we have lost in data?”
Intro

In order to go from data to information, to
knowledge and to wisdom,
we need to reduce the complexity of the data.
Complexity can be reduced on
- case level : cluster analysis
- on variable level: factor analysis
Intro

Cluster analysis can get you from this:
To this:
What cluster analysis does
a
b
d
e
f
c

Cluster analysis
• generate groups which are similar
• homogeneous within the group and as much as
possible heterogeneous to other groups
• data consists usually of objects or persons
• segmentation based on more than two variables

Cluster analysis
• generates groups which are similar
• the groups are homogeneous within themselves
and as much as possible heterogeneous to other
groups
• data consists usually of objects or persons
• segmentation is based on more than two
variables

Examples for datasets used for cluster
analysis:
• socio-economic criteria: income, education, profession,
age, number of children, size of city of residence ....
• psychographic criteria: interest, life style, motivation,
values, involvement
• criteria linked to the buying behaviour: price range, type
of media used, intensity of use, choice of retail outlet,
fidelity, buyer/non-buyer, buying intensity

Proximity Measures
• Proximity measures are used to represent the nearness of two
objects
• relate objects with a high similarity to the same cluster and objects
with low similarity to different clusters
• differentiation of nominal-scaled and metric-scaled variables
m
d(yi,ys) = [∑ |yij-ysj|r
]1/r
j=1
y = vector
i,s = different objects
j = the different characteristics
r = changes the weight of assigned distances
the calculation of the distances measures is the basis of the
cluster analysis.

Two phases:
1. Forming of clusters by the chosen data set – resulting
in a new variable that identifies cluster members
among the cases
2. Description of clusters by re-crossing with the data

Cluster Algorithm in agglomerative hierarchical
clustering methods – seven steps to get clusters
1. each object is a independent cluster, n
2. two clusters with the lowest distance are merged to
one cluster. reduce the number of clusters by 1 (n-1)
3. calculate the the distance matrix between the new
cluster and all remaining clusters
4. repeat step 2 and 3, (n-1) times until all objects form
one reminding cluster

Finally…
1. decide upon the number of clusters you want to keep
(decision often based on the size of the clusters)
2. description of the clusters by means of the cluster-
forming variables
3. appellation of the clusters with catchy titles

Cluster 5Cluster 4Cluster 3Cluster 2Cluster 1

Practical Example
Consumers and Fair Trade Coffee (1997!)
214 interviews of consumers of fair trade
coffee (personal and telephone interviews)
Cluster analysis in order to identify consumer
typologies
Identification of 6 clusters
Description of these clusters by further
analysis: comparison of means, crosstabs etc.

Consumers and Fair Trade Coffee
Description of clusters:
Cluster 1 (11,6%): “self-oriented fair trade buyer”
Cluster 2 (13,6%): “less ready to take personal
constraints”
Cluster 3 (18,2%): ”less engaged about fair trade”
Cluster 4 (32,2%): “intensive buyer”
Cluster 5 (18,7%): “value-oriented”
Cluster 6 (5,6%): “does not like the taste of fair trade
coffee”

Description of Cluster 1 (11,6%): “self-oriented fair trade
buyer” :
Searches satisfaction by doing the good thing
Is not altruistic
Buys occasionally
Sticks to his conventional coffee brand
High level of formal education
Frequently religious (catholic or protestant)

Description of Cluster 2 (13,6%): “less ready to take
personal constraints”
States that “fair trade coffee is hard to find”
Feels responsible for fare development issues
Believes that fair trade is efficient for developing
countries
Is less ready to go to special fair trade outlets
Buys conventional coffee
Likes the taste of fair trade coffee

Description of clusters Cluster 3 (18,2%): ”less engaged
about fair trade” :
Feels no personal responsibility with regard to
development questions
Doesn’t see the efficiency of the consumption of fair
trade goods
The only thing that can make him change is the
influence of friends
Is older then the average fair trade buyer and has less
formal education

Description of clusters: Cluster 4 (32,2%): “intensive
buyer”
Has abandoned conventional coffee brands
Has started to buy fair trade quite a while ago (> 3
years)
Shops frequently in fair trade stores (and not in organic
retail)
Is ready to act for fair development and talks to friends
about it
Relatively young, with low incomes and high
educational values

Description of clusters: Cluster 5 (18,7%): “value-
oriented”
Together with cluster 4 highly aware of development
issues
Ready to act and to constraint consumption habits
Buys for altruistic reasons
Highly involved in social / political action
Most frequently women, highest household income
among all clusters
Own security is the basis for solidary action

Description of clusters: Cluster 6 (5,6%): “does not like
the taste of fair trade coffee”
Lowest purchase intensity of all clusters
Not willing to accept constraints in consumption habits
or higher prices
Most members of these group are attached to a
conventional coffee brand
Relatively high incomes, age within the average of all
groups, lower level of formal education
Less religious than other groups.

Conclusion /
discussion
Advantages
• no special scales of measurement necessary
• high persuasiveness and good assignment to realisable
recommendations in practice
Disadvantages
• choice of cluster-forming variables often not based on
theory but at random
• determination of the right number of clusters often time-
consuming – often decided upon arbitrarily
• high influence on the interpretation of the scientist, difficult
to control (good documentation is needed)

Russell .L. Ackoff, "From Data to Wisdom," Journal of Applied Systems
Analysis 16 (1989): 3-9.
Milan Zeleny, "Management Support Systems: Towards Integrated
Knowledge Management," Human Systems Management 7, no 1
(1987): 59-70.
Tashakkori, A. and Ch. Teddlie: Combining Qualitative and Quantitaive
Approaches. Applied Social Research Methods Series, Volume 46.
Thousand Oaks, London, New Delhi, 1998.
xyxy
Sources

Online data science course

More Related Content

Recently uploaded (20)

Featured (20)

Online data science course