Why Naive Bayes works

Case 1. Categorical features and frequencies

Implementation with RubixML

Here the same case is implemented with RubixML: we train Naive Bayes on the same data and see which class the model assigns to a new sample.

Example of use


                
<?php

use Rubix\ML\Datasets\Unlabeled;

include __DIR__ . '/rubix-code.php';

// New sample to classify (must use the same categorical encoding as during training).
$sample = ['from_ads', 'has_account'];

// RubixML expects samples to be wrapped into a Dataset object.
$dataset = new Unlabeled([$sample]);

// Predict class label.
$prediction = $model->predict($dataset);

// Print the predicted label.
print_r($prediction);

Result: Memory: 0.366 Mb Time running: 0.009 sec.


                Array
(
    [0] => buyer
)

RubixML returns the predicted class label for the given sample. Under the hood it applies the same Naive Bayes idea (class prior × conditional likelihoods), so the result should match the pure PHP implementation for the same data.