Data Mining Assignment: Association and Clustering

This data mining assignment analyzes cosmetic data using XL Miner to extract association rules with a minimum confidence level of 50% and then 75%. It further explores hierarchical clustering and K-Means clustering techniques to segment customers based on their flight transaction history. The assignment includes dendrogram interpretation, cluster labeling, and targeted offer suggestions for different customer segments.

DATA MINING
[Pick the date]
Student id

Table of Contents
Question 1........................................................................................................................................1
Question 2........................................................................................................................................4
Reference.........................................................................................................................................5

Question 1
Association Rule
The association rules have been predicted for the given cosmetic data using XL Miner as an aid.
In this process, the minimum confidence level to be considered would be 50%.
(i) The three rules in relation to the association rules output which has been generated above
are highlighted below. The interpretation of these rules is based on definition of the rule
which is provided at the top of the association output.

(ii) There is redundancy of rule when no value addition or incremental value is added by
the given rule. In such a scenario, the rule which is inferior would be discarded on
account of redundancy so as to bring about an improvement in the overall output. For
the association rule output, it can be concluded that potential redundancy is observed
for rule 2 when seen in the context of rule 1 (Abramowics, 2013).
The overall importance of a given rule is reflected by the lift ratio and also the rules
in the output are arranged in the decreasing order of lift ratio. Further, utility of the
rules also is expressed through the use of confidence level. These rules when viewed
in the light of the above two rules tend to highlight significant information in relation
to the customer purchasing behavior which have immense utility in understanding the
customer preferences and the manner in which they do purchases (Ragsdale, 2014).
(iii) The earlier output which has been highlighted was based on the minimum confidence
level of 50%. Hence, the critical observation in the above case displayed only those
rules where confidence level was higher than 50%. But, in this question there has
been an increase in the minimum confidence level from the level of 50% to 75%.
XLMiner output

The output above clearly suggests that only one rule is being displayed with a confidence
level of 100%. The other rules are not being displayed since the corresponding level
would be lesser than 75% and hence not apparent in the output above. Thus, the number
of rules displayed have declined with increasing minimum confidence level.

Question 2
Cluster Analysis
(a) Dendogram is a critical output which is available for hierarchical clustering and is
highlighted as follows. It tends to aid in the cluster formation process based on the
underlying distance.
:
In the above case, a line which is slightly lower than 1,000 say 999 would lead to the three
cluster formation which was originally envisioned while running the hierarchical clustering.
(b) The normalisation of data is critical and the absence of this would lead to certain
problems which are listed as follows (Shumueli et. al., 2016).
The distance computation would not be accurate as the scale effect would be observed
especially when unequal weights are used for this process.

The accuracy of the model would be hampered as the variables having higher scale would
distort the cluster formation.
c) The cluster labeling needs to be carried out taking into consideration the respective centroid
distance and the common traits. Through this process, the cluster labeling is done below.
Cluster 1 – Termed as “Middle Class Flyers” owing to the low flight transactions along with the
moderate spending levels.
Cluster 2 – Termed as “High Networth Flyers” owing to the very high flight transactions during
the last one year coupled with other indicators quite high.
Cluster 3: Termed as “Infrequent Flyers” owing to the high non-flight transactions but low flight
transactions.
d) Another clustering technique which has been applied on the East West Airlines data is the K
Means Clustering. The relevant output for the same is reflected below.
It is apparent from the above output obtained that this is significantly different from the output
indicated in part (a) and (b). This may be reflected by taking any particular cluster. For example
cluster 1 is chosen. Based upon the measures, a noteworthy fact is that the flight transactions
seem very high which is a key characteristic of the flyers who have high networth. This cluster
classification is different from that experienced in hierarchical clustering and hence it can be
easily reflected that the a similar pattern does not from in the two clustering techniques
(Ragsdale, 2014).

e) The requisite cluster targeting with suitable offers is shown below.

Reference
Abramowics, W. (2013) Business Information Systems Workshops: BIS 2013 International
Workshops (5th ed.). New York: Springer.
Ragsdale, C. (2014) Spread sheet Modeling and Decision Analysis: A Practical Introduction to
Business Analytics (7th ed.). London: Cengage Learning.
Shumueli, G., Bruce, C.P., Yahav, I., Patel, R. N., Kenneth, C., & Lichtendahl, J. (2016) Data
Mining For Business Analytics: Concepts Techniques and Application (2nd ed.).London:
John Wiley & Sons.