Reputation: 744
I was reading this webpage which was describing an example of how Naive Bayes classification works:
http://www.statsoft.com/textbook/naive-bayes-classifier
Their example involves looking at the probability of a class given other points in the sample's vicinity.
This looks like some sort of hybrid of naive bayes and k-Nearest-Neighbors. Is this true and is this truly the naive bayes algorithm, or is this some new algorithm?
Thanks.
Upvotes: 2
Views: 104
Reputation: 2272
The example is not how a naive bayes classifier work in any way, how ever it's not taken from thin air.
Under certain assumptions (like that the naive bayes is a perfect model for the data and infinite amount of data) we have that the
"number of COLOR in vicinity X/number of COLOR"
is the
"density of COLOR near X"
which is proportional to p(x|COLOR)
so finally p(COLOR|x) <proportional to> p(x|COLOR)p(COLOR)
It shows a relation between the kNN algorithm and naive bayes but nothing more.
Upvotes: 1
Reputation: 1270
I don't think it is any other algorithm. It is clearly naive Bayes classifier. He took some possible example to explain the concept.
Upvotes: 0