E
E-
-C
CO
OM
MM
ME
ER
RC
CE
E
–
–
I
IT
T4
43
30
0
V
VU
U
© Copyright Virtual University of Pakistan 144
Lesson 35
CONFIDENCE AND SUPPORT
There are two terms/measures used in association, that is, support and confidence. Confidence’ is a
measure of how often the relationship holds true e.g, what percentage of time did people who bought milk
also bought eggs. Support means what is the percentage of two items occurring together overall.
Mathematically, they can be expressed as follows if we take the example of eggs and milk:
Confidence = Transactions (eggs+milk)
Transactions (eggs or milk or both)
In case no. of transactions involving eggs and milk are 25 and those involving eggs or milk or both are 75
then confidence is 25/75*100=33.3%
Support = Transactions (eggs+milk)
Total no. of transactions
In case no. of transactions involving eggs and milk are 10 and total no. of transactions in a day are 50 then
support is 10/50*100 = 20%
Suppose if confidence is 90% but the support is 5%., then we can gather from this that the two items have
very strong affinity or relationship with each other such that when an item is sold the other is sold together,
however, the chance of this pair being purchased out of the total no. of transactions is very slim, just 5%.
One can adjust these measures to discover items having corresponding level of association and accordingly
set marketing strategy. So, if I feed the data to the association mining tool and specify the percentage of
confidence and support, it will list down the items that have association corresponding to these percentages.
Results of association mining are shown with the help of double arrows as indicated below:
Bread Å----Æ Butter
Computer Å----Æ Furniture
Clothes Å----Æ Shoes
Using the result of association mining, a marketer can take a number of useful steps to set or modify
marketing strategy. For example, items that have closeness/affinity with each other can be shelved together
to improve customer service. Certain promotional schemes can be introduced in view of the association
mining result etc.
Characterization
It is discovering interesting concepts in concise and succinct terms at generalized levels for examining the
general behavior of the data. For example, in a database of graduate students of a university the students of
different nationalities can be enrolled in different departments such as music history, physics etc. We can
apply characterization technique to find a generalized concept/answer in response to the question that how
many students of a particular country are studying science or arts. See the following example:
Student name Department City of residence
Imran History Karachi
Alice Physics London
Ali Literature Lahore
Bob Mathematics Toronto
…
In the above example, characterization tool can, for that matter, tell us that 02 Pakistani students are
studying arts. Note that the concept of location and the field of education are generalized to Pakistan and
arts, respectively.
The two algorithms used in characterization are Version Space Search and Attribute-Oriented Induction.
docsity.com
