MSCIT 521' In class exercise

1
MSCIT 521. In class exercise

• Sept 18, 2008

2
Question 1

• Give an example of a dataset and a classifier
  such that applying Leave-One-Out classification
  on this dataset will give 0 accuracy.

3
Question 3.

• Apply Bootstrapping on the weather dataset, by
  using a random number generator such as
  http//www.random.org/integers/
• What is the training data and what is the test
  data?

4
Question 4

• Suppose we are searching for a document related
  to a person (e.g., your instructor).
• Select two search engines
• What is the precision and recall for each search
  engine, based on first result page?
• Assume that there are 100 documents relevant to
  your instructor
• Which one is more accurate?

5
Question 5

• What is the entropy of our class?
• In terms of what attributes?
• Which has the higher value?

6
Question 6 Ranking

• Think of a rule to predict whether a classmate in
  our class likes classical music (with values T
  or F), that can generate probability of Pr(Tx),
  where x is the attributes of the classmate.
• Rank the students in our class based on the
  probability values
• Plot the Lift Chart.