Naïve Bayesian Sample Clauses

Naïve Bayesian. Naïve Bayesian [92] is well-known for its scalability: both training and testing have time complexity O(n). To adapt the algorithm for online learning, the training step needs to be modified to update the count of classes and attributes after each instance is processed. The probability of each class p (Ck)and the probability of p (xi | Ck) are updated accordingly (xi is the value of input vector X = (x1, x2, x3, ……, xi) ). These updates can be made in time O(1), hence the overall modified algorithm remains O(n). To predict output, the original prediction function of Naïve Bayesian can be applied directly in constant time: prediction = arg max P(Ck)∏P(xi | Ck) To handle skewed data, especially when costs of misclassification for different classes are not equivalent, bias may be applied to p (Ck) of the prediction function so as to prioritize the preference of different classes.