Naive Bayes

Question 1

What topics are covered in the outline of this deck?

Answer 1

Bayes Theorem
Bayesian Classification
Naibe Bayes for Classification

Question 2

What is the mathematical formula of Bayes’ formula?

Answer 2

P(A|B) = [P(B|A) * P(A)] / P(B)

P(A|B) = Probability of A given B (the posterior probability)
P(B|A) = Probability of B given A (the likelihood)
P(A) = Prior probability of A
P(B) = Probability of B (marginal probability)

Question 3

In Bayesian Classification what do we do when we have many examples for each possible combination of values for all predictor attributes, but different classes?

Answer 3

The optimal classification, to maximise accuracy, would be to choose the most frequent class for that combination of attribute values.

Question 4

What are some limitations of Naïve Bayes?

Answer 4

Assumes each attribute is independent from all other attributes, given a class.
- This is usually an unrealistic assumption
- This is also why Naïve Bayes is naïve

Classification accuracy is sensitive to redundant attributes

Question 5

What are some strengths of Naïve Bayes

Answer 5

Is able to deal with missing attributes.
- Missing attributes can be ignored because attributes are conditionally independent given the class