Naive Bayes §

---

Notes and Ideas: §

• Text classification
  • naive bayes is easy because we make a very strong(and wrong) assumption.
• Ex.
  • Basic spam classifier:
    • use proportion of spam from training set as prediction
    • P(spam|X) where X is our features from our emails -> logistic regression
    • prior information ( training set )

• Bayes Theorem

  Bayes §

  ---

  Example: From the prior information: we would have the Truepositive rate (sensitivity) = 0.85 = P(+ | D(true)). P(- | D(true)) = 0.15

  Hence: P(- | D(false)) = 0.1, P(+ | D(false)) = 0.1 P(D) = 0.0001

  Goal: find P(D | +): test positive and the true prior information:

  ---

  Theorem:

  1. =

  ---

  Definition:

  Link to original

• =
• x = []

---

methods: §

Bernoulli method: §

• each features

---

TAGS §

#classification