Naive Bayes

Question 1

Naive Bayes

Answer 1

Conditional Probability
P(C|A) = P(A|C)P(C)/P(A)
x’ = 
class = arg max P(y|x’)
P(y|x’) = P(x’|y) * P(y) / P(x’)
Normalization

Question 2

Zero Frequency Problem

Answer 2

If value does not occur. Probability = 0
Posteriori = 0
We add 1 to the count. (Laplace)
0r we can add a constant different. (2+ u/3) / (9+u)

Question 3

Missing Values

Answer 3

Not included in freq count.

Testing: omitted from calculation

Question 4

Numeric Attributes

Answer 4

Assume that attributes have a normal or Gaussian probability
Sample mean: mean = (1/n)sum(xi)
std = sqrt ((1/(n-1)sum(xi-mean)^2))
density function: gauss

Question 5

Bayesian Belief Networks

Answer 5

Allows us to specify which pair of attributes are conditionally independent
Graphical representation of probabilistic relationships among a set of random variables
Direct acyclic graph
Probability table