programming-from-a-to-z / bayes-classifier-js Goto Github PK

There's a function called probabilities() which calculates everything. The library should probably execute this function internally after each call to train()? Also, naming-wise it should probably be called something else. .. calculate()?

https://github.com/NaturalNode/natural#classifiers

https://github.com/ttezel/bayes

Some ideas:

• Highest probability category
• List of all categories and their probabilities
• List of words and their scores / probabilities?

What else?

This is what I'm doing now:

if (word[category].prob < 0.01) word[category].prob = 0.01;
if (word[category].prob > 0.99) word[category].prob = 0.99;

Allow user to pass in their own function for token validation?

• train based on user input?
• train based on preloaded text files?

Allow use to pass in their own function to tokenize the text?