Frequency Based Classification Algorithms_ important
0 likes5 views
This document discusses two types of frequency-based classification methods: K-Nearest Neighborhood (KNN) and Naive Bayes. KNN is a simple counting-based method that measures distances between data points to classify them, but can be ad-hoc. Naive Bayes uses Bayes' theorem to calculate conditional probabilities of class membership given attribute values in order to classify data points into classes. It makes the assumption that attributes are conditionally independent given the class.
1 of 5
Download to read offline
More Related Content
More from HaritikaChhatwal1 (12)
Ad
Recently uploaded (20)
Ad
Frequency Based Classification Algorithms_ important
- 2. Types • K Nearest Neighborhood (KNN) • Naïve Bayes
- 3. K-Nearest Neighbour • Simplest: Count Based • Not sensitive to variance-covariance • Measure ‘distances’ • Distances ‘type’ • Very useful and simple • But bit ad-hoc • Other applications 12-Dec-17 3
- 4. Naïve Bayes • Bayes Theorem • 𝑃 𝑦𝑖 = 1|𝑋 = 𝑃 𝑦𝑖=1 ∗𝑃 𝑋|𝑦𝑖=1 𝑃 𝑦𝑖=1 ∗𝑃 𝑋|𝑦𝑖=1 +𝑃 𝑦𝑖=0 ∗𝑃 𝑋|𝑦𝑖=0 • Re visit Bayes Theorem with an Example: • P(G)=0.95, P(a=g|G)=0.9, P(a=g|B)=0.15 • So what is P(G|a=g)?
- 5. Naïve Bayes • How does it Work • No a priori specification is needed • Can be used to do multi class classification as well