Getting Started with Ruby Classifier

This tutorial will get you up and running with the classifier gem in under 5 minutes. By the end, you’ll have a working text classifier.

Prerequisites

• Ruby 3.1 or higher
• Bundler (optional, but recommended)

Installation

Install the gem directly:

gem install classifier

Or add it to your Gemfile:

gem 'classifier'

Then run:

bundle install

Your First Classifier

Let’s build a simple spam detector. Create a file called spam_detector.rb:

require 'classifier'

# Create a classifier with two categories
classifier = Classifier::Bayes.new 'Spam', 'Ham'

# Train it with some examples
classifier.train(spam: "Get rich quick! Buy now!")
classifier.train(spam: "You've won a million dollars!")
classifier.train(spam: "Click here for free stuff")

classifier.train(ham: "Meeting tomorrow at 10am")
classifier.train(ham: "Please review the attached document")
classifier.train(ham: "Thanks for your email")

# Now classify some new text
puts classifier.classify "Claim your free prize today!"
# => "Spam"

puts classifier.classify "See you at the meeting"
# => "Ham"

Run it:

ruby spam_detector.rb

Understanding the Output

The classify method returns the most likely category for the given text. Under the hood, the Bayesian classifier:

1. Tokenizes the text into words
2. Stems each word to its root form
3. Calculates the probability of each category
4. Returns the category with the highest probability

Getting Probability Scores

Want to see the actual scores? Use classifications:

scores = classifier.classifications "Limited time offer!"
puts scores
# => {"Spam" => -5.2, "Ham" => -9.8}

Higher (less negative) scores indicate higher probability. The classifier returns log probabilities to avoid numerical underflow with large datasets.

Next Steps

Now that you have a basic classifier working, explore these topics:

• Build a Complete Spam Filter - A production-ready email classifier
• Bayes Basics Guide - Deep dive into how Bayesian classification works
• LSI for Semantic Search - Find related documents using meaning, not just keywords
• kNN Classification - Instance-based classification with interpretable results
• TF-IDF Vectorizer - Transform text into weighted feature vectors

Quick Reference

# Create classifier
classifier = Classifier::Bayes.new 'Category1', 'Category2'

# Train with keyword arguments (recommended)
classifier.train(category1: "example text")

# Batch train with arrays
classifier.train(category1: ["text 1", "text 2"], category2: ["text 3"])

# Legacy APIs (still work)
classifier.train :Category1, "example text"
classifier.train_category1 "example text"

# Classify
result = classifier.classify "text to classify"

# Get scores
scores = classifier.classifications "text to classify"