Question 2

1. (i) Explain how Association rules are used in the Business Intelligence (BI) context. You may use appropriate examples to illustrate. (ii) In addition to confidence and support, some other measures used to describe the interestingness of association rules are: lift, leverage and conviction.

An extract of the result of the Apriori algorithm on the classic zoo dataset from Weka is given:

venomous=false	tail=true	71	==> backbone=true	71
lift:(1.22)	lev:(0.13)	[12]	conv:(12.65)>

Explain the meaning of lift and explain how to calculate this measure. Use any example to illustrate.

3. Generally, we will be more interested in association rules having high confidence. However, association rules having 100% confidence might not be as interesting as association rules with 99% confidence. Explain why we would prefer rules with 99% confidence and what they might indicate.

b. i. A Clustering algorithm is a commonly used techniques for insight creation in Business Intelligence. List and explain the different steps of the K-means Clustering algorithm.

2. Use examples to illustrate how we can measure the similarity or dissimilarity between non-numerical objects for clustering in a Business Intelligence context.