Skip to content

[source]

Naive Bayes#

Categorical Naive Bayes is a probability-based classifier that uses counting and Bayes' Theorem to derive the probabilities of a class given a sample of categorical features. The term naive refers to the fact that Naive Bayes treats each feature as if it was independent of the others even though this is usually not the case in real life.

Note

Each partial train has the overhead of recomputing the probability mass function for each feature per class. As such, it is better to train with fewer but larger training sets.

Interfaces: Estimator, Learner, Online, Probabilistic, Persistable

Data Type Compatibility: Categorical

Parameters#

# Name Default Type Description
1 priors null array The class prior probabilities as an associative array with class labels as keys and their prior probabilities as values totalling 1. If null, then priors will automatically be computed from the training data.
2 smoothing 1.0 float The amount of Laplace smoothing added to the probabilities.

Example#

use Rubix\ML\Classifiers\NaiveBayes;

$estimator = new NaiveBayes(2.5, [
    'spam' => 0.3,
    'not spam' => 0.7,
]);

Additional Methods#

Return the class prior probabilities:

public priors() : float[]|null

Return the counts for each category per class:

public counts() : array[]|null


Last update: 2021-03-27