DigikalaNext contest - 2019
Implemented a comment verification system using CountVectorizer, Bayes algorithm and Digikala website's comments dataset
-
Scikit-learn's Countvectorizer :
Transform a given text into a vector on the basis of the frequency (count) of each word that occurs in the entire text.
-
Bayes theorem
The “prior” P(A) and the “evidence” P(B) are the probabilities of observing A and B independently in the document, whereas the “posterior” and the “likelihood” are the conditional probabilities of observing A given B and vice versa.
In this project what we are going to find is this:
While x is a feature vector containing the sequence of words in the given comment.
The “Naive” assumption that the Naive Bayes classifier makes is that the probability of observing a word is independent of others. Therefore, the probability of that comment being a spam is the product of seeing each of the words in the comment if a spam comment.