bow_mnb

Given training data in BOW format, trains a multinomial Naive Bayes classifier

One method for training and testing, optional parameters - one line of code to train and test the classifier

Usage example: First download and install the package in your terminal:

$python3 -m pip install bow_mnb

Then in your python file:

import bow_mnb.mnb as mnb
mnb.train_and_test('./training/files/', './validation/files/')

Specify paths for training and testing data as folders containing subfolders, each with the name of their class label

Example:

Class labels A, B, and C

Directory structure should be as follows:

training
    ├──A
       ├──doc1.txt
       ├──doc2.txt
       └──...
    ├──B
       ├──doc1.txt
       ├──doc2.txt
       └──...
    └──C
       ├──doc1.txt
       ├──doc2.txt
       └──...

For the above directory structure, I would train the classifier as follows:

mnb.train_and_test('./training/')

An individiual document should be in BOW format (i.e., each line is a given word and its number of occurences, followed by a newline character)

Example:

snow 5
orange 2
the 98
blog 4
...

currently no support for non-BOW format // TO-DO - more file type and format handling

Optional paramters: set fileType to be txt or res, default is txt set tfidf=True to use TF-IDF vectorization instead of count vectorization (default) set stem=True to use SnowballStemmer on tokens try different laplace value to tune classifier (0.1, 0.05, 0.01, 0.005, 0.001, etc.) set outputFile=True to write predicted labels for some test or validation data (per document) to a single text file