naiveBayes Method

The naiveBayes() method implements a Naive Bayes classifier that predicts a categorical label based on given features. It returns a classifier function that can be used to predict labels for new items.

Basic Syntax

naiveBayes<K extends keyof T>(
  features: K[],
  label: K
): (item: Pick<T, K>) => T[K]

Examples

Basic Usage

import { collect } from 'ts-collect'

const data = collect([
  { color: 'red', size: 'large', category: 'A' },
  { color: 'blue', size: 'small', category: 'B' },
  { color: 'red', size: 'small', category: 'A' }
])

const classifier = data.naiveBayes(
  ['color', 'size'],
  'category'
)

const prediction = classifier({ color: 'red', size: 'small' })
console.log(prediction) // 'A'

Feature Classification

interface Document {
  words: string[]
  language: string
}

const documents = collect<Document>([
  { words: ['hello', 'world'], language: 'english' },
  { words: ['bonjour', 'monde'], language: 'french' },
  { words: ['hola', 'mundo'], language: 'spanish' }
])

const languageClassifier = documents.naiveBayes(
  ['words'],
  'language'
)

Real-world Example: E-commerce Category Prediction

interface Product {
  name: string
  price: number
  description: string
  brand: string
  category: string
}

class ProductCategorizer {
  private trainingData: CollectionOperations<Product>
  private classifier: (item: Pick<Product, 'price' | 'brand'>) => string

  constructor(products: Product[]) {
    this.trainingData = collect(products)
    this.classifier = this.trainingData.naiveBayes(
      ['price', 'brand'],
      'category'
    )
  }

  predictCategory(product: Pick<Product, 'price' | 'brand'>): string {
    return this.classifier(product)
  }

  batchPredict(products: Array<Pick<Product, 'price' | 'brand'>>) {
    return products.map(product => ({
      ...product,
      predictedCategory: this.predictCategory(product)
    }))
  }
}

// Usage example
const categorizer = new ProductCategorizer([
  {
    name: 'Pro Laptop',
    price: 1299,
    description: 'High-performance laptop',
    brand: 'TechCo',
    category: 'Electronics'
  },
  {
    name: 'Office Chair',
    price: 299,
    description: 'Ergonomic chair',
    brand: 'FurnishCo',
    category: 'Furniture'
  },
  {
    name: 'Wireless Mouse',
    price: 49,
    brand: 'TechCo',
    description: 'Bluetooth mouse',
    category: 'Electronics'
  }
])

const newProduct = {
  price: 1199,
  brand: 'TechCo'
}

console.log(categorizer.predictCategory(newProduct))

Return Value

• Returns a classifier function
• Function accepts partial objects with feature keys
• Returns predicted label of same type as training data
• Maintains type safety through generics
• Handles missing feature values
• Uses Laplace smoothing for probability calculation

Common Use Cases

1. Product Categorization

   • Automatic category assignment
   • Product classification
   • Inventory organization
   • Tag prediction

2. Customer Analysis

   • Customer segmentation
   • Behavior prediction
   • Preference classification
   • Target group identification

3. Content Classification

   • Text categorization
   • Document classification
   • Language detection
   • Topic identification

4. Quality Control

   • Defect prediction
   • Quality classification
   • Automated sorting
   • Issue categorization

5. Risk Assessment

   • Fraud detection
   • Risk classification
   • Order validation
   • Security screening