Naive Bayes Model is based the assumption that the variables {π1, β¦, ππ} areΒ conditionally independentΒ givenΒ π
- π(π|π1, β¦,Β ππ) =Β π(π1|π) β¦Β π(ππ|π)Β π(π) /Β π(π1, β¦,Β ππ)
- π(π|π1, β¦,Β ππ) βΒ π(π1|π) β¦Β π(ππ|π)Β π(π)
if the variablesΒ {π1, β¦, ππ} happen to NOT be conditionally independent givenΒ π, this would cause an Overcounting Problem
In other words, correlated variables in {π1, β¦, ππ} which breaks conditional independence