I am attempting to use a Naive Bayes classifier to classify text. To accomplish this I have created an Excel sheet with a binary distribution for three variables. The workbook can be found here. Assuming that my math is correct, my questions are:
- Can my training set can be expanded as I classify new inputs? In other words, every time I check a classification that the model has produced, I can add it to the training model but then I might have an uneven number of examples for each class. Is this a problem?
- How can I incorporate a prior distribution to the equation? If I for example know from prior data that Class A is twice as likely than Class B?
- How can I incorporate tf–idf to the equation? I can analyze all the data sets a priori and the frequencies of each word in both the corpus and each document, but am unsure how to incorporate this into the Classifier.
Thanks in advance for everyone help.