Struct NaiveBayesClassifier 

pub struct NaiveBayesClassifier { /* private fields */ }

Multinomial Naïve Bayes text classifier with Laplace (additive) smoothing.

Internally works in log-space to prevent floating-point underflow.

Example

use scirs2_text::text_classification::NaiveBayesClassifier;

let mut clf = NaiveBayesClassifier::new(1.0);
let texts  = &["spam spam buy now", "hello friend good morning"];
let labels = &["spam", "ham"];
clf.fit(texts, labels).unwrap();
assert_eq!(clf.predict("buy now cheap").unwrap(), "spam");

Implementations

impl NaiveBayesClassifier

pub fn new(alpha: f64) -> Self

Create a new classifier with smoothing parameter alpha.

alpha = 1.0 is the standard Laplace (add-one) smoothing.

Errors

Returns TextError::InvalidInput when alpha <= 0.

pub fn fit(&mut self, texts: &[&str], labels: &[&str]) -> Result<()>

Train the classifier.

Errors

Returns TextError::InvalidInput when:

• texts and labels have different lengths.
• The corpus is empty.
• alpha <= 0.

pub fn predict(&self, text: &str) -> Result<String>

Predict the most probable class for text.

Errors

Returns TextError::ModelNotFitted if fit has not been called.

pub fn predict_proba(&self, text: &str) -> Result<Vec<(String, f64)>>

Return (class_name, probability) pairs sorted by probability descending.

Probabilities are derived from log-scores via softmax normalisation.

Errors

Returns TextError::ModelNotFitted if fit has not been called.

pub fn score(&self, texts: &[&str], labels: &[&str]) -> Result<f64>

Compute accuracy on a labelled test set.

Errors

Returns TextError::InvalidInput when lengths differ.