Data Mining - Association Rules

Question 1

What are itemsets in association rules?

Answer 1

Possible combinations of items. Not necessarily what people buy.
-> It can also be a single item.

Question 2

How are Rules represented?

Answer 2

IF-THEN format.

If -> antecedent
Then -> consequent

Question 3

Which two steps are there in association rules process

Answer 3

1. Generation of Frequent Item sets (apriori)

2. Selecting strong rules

Question 4

What are frequent item sets?

Answer 4

You can not check all combinations of items. Therefore, we only look at the frequent item sets.

Question 5

What is the criterion for frequent item sets?

Answer 5

Support

Support(A and C) = frequency (A and C)

Easy computation: you look in your table how often that itemset is there in the records.

• You can divide by the total number of records in the table to get a percentage

Question 6

How does the process of generating the most frequent item set go?

Answer 6

You set a minimum suport.
You check for single item item sets. above that support.
Then you check for double item item sets above that support and so on.

Question 7

How do you measure how strong those associations are?

Answer 7

1. Support
2. Confidence
3. Lift

Question 8

How do you calculate confidence?

Answer 8

Frequency A and C happen / Frequency A happens

Question 9

How do you calculate lift?

Answer 9

Confidence / (frequency C / n)

Question 10

What does a lift of > 1 indicate?

Answer 10

There is a dependent relation between the antecedent and consequent

Question 11

What does a lift of <1 indicate?

Answer 11

The presence of the antecedent has a negative effect on the presence of the consequent.

Question 12

What does a lift of approx 1 indicate?

Answer 12

The antecedent and consequent are independent.