BAYESIAN CLASSIFIER

Simple bayesian classifier using naturaljs

Train

node index.js train

The train command needs a training set. By default, the program looks for a training_set.json which contains labeled samples.

{
  "label1": [
    "sample1",
    "sample2",
    "sample3"
  ],

  "label2": [
    "sample1",
    "sample2",
    "sample3"
  ]
}

To load a different training set specify the file name on the command line:

node index.js train -t "my_training_set.json"

A model is created and a file is saved in the current directory with (default) name model.json. To create a model with a different name:

node index.js train -m "my_model.json"

It is possible to combine the options to load some other training set and to save the model to some other file:

node index.js train -t "my_training_set.json" -m "my_model.json"

At any moment it is possible to access a simple help screen:

node index.js train --help

Classify

node index.js classify

The classify command loads a model file (default model.json) and tries to classify one or more test samples.

node index.js classify "My sample that needs to be classified"

When run without a sample argument, it will read samples from a JSON file (default samples.json).

[
  "sample1",
  "sample2",
  "sample3"
]

Run the classify command without arguments:

node index.js classify

To specify a different test samples file:

node index.js classify -s "my_test_samples.json"

It is also possible to specify a different model file to use:

node index.js classify -m "my_model.json"

The classification will show how the test samples score against each trained label. In order to show only the best match:

node index.js classify -b

Options can be combined and a help screen is available

node index.js classify --help