Given a set of testing data, with a bag of words encoding on each email message. We end up with a matrix of messages and their features, as binary encoded values. That is, for each email message, a feature x will be encoded to a 1 if it is present in an email message. Programmed classifiers (neural network, naives bayes, logistic regression and linear regression) from scratch to go through 54 one hot encoded features per input and categorize it.
