Overview
LabelEncoder encodes labels as consecutive integers from 0 to n_classes-1. This is useful for converting categorical labels to numeric format for algorithms that require numeric input.
Constructor
No parameters required.
Properties
Sorted unique classes found during fitting.
Methods
fit
Fit label encoder to a set of labels.
Target labels array. Must be a non-empty array of finite numbers.
this - The fitted encoder instance.
transform(y: Vector): Vector
Target labels to encode. All labels must have been seen during fit.
Vector - Encoded labels.
fitTransform(y: Vector): Vector
fit(y).transform(y).
Target labels to fit and encode.
Vector - Encoded labels.
inverseTransform(y: Vector): Vector
Encoded labels (integers). All values must be valid indices.
Vector - Original labels.
Example
import { LabelEncoder } from 'bun-scikit';
// Original labels
const labels = [1, 2, 2, 6, 1, 6, 6, 2];
// Create and fit encoder
const encoder = new LabelEncoder();
const encoded = encoder.fitTransform(labels);
console.log('Classes:', encoder.classes_);
// Output: Classes: [1, 2, 6]
console.log('Encoded:', encoded);
// Output: Encoded: [0, 1, 1, 2, 0, 2, 2, 1]
// 1 -> 0, 2 -> 1, 6 -> 2
// Decode back to original labels
const decoded = encoder.inverseTransform(encoded);
console.log('Decoded:', decoded);
// Output: Decoded: [1, 2, 2, 6, 1, 6, 6, 2]
Example: Classification Pipeline
import { LabelEncoder } from 'bun-scikit';
// Original string-like numeric labels
const yTrain = [10, 20, 30, 20, 10, 30];
const yTest = [20, 10, 30];
// Encode training labels
const encoder = new LabelEncoder();
const yTrainEncoded = encoder.fitTransform(yTrain);
console.log('Training labels encoded:', yTrainEncoded);
// Output: Training labels encoded: [0, 1, 2, 1, 0, 2]
// Encode test labels (must use same encoder)
const yTestEncoded = encoder.transform(yTest);
console.log('Test labels encoded:', yTestEncoded);
// Output: Test labels encoded: [1, 0, 2]
// After prediction, decode back
const predictions = [0, 1, 2, 0]; // Model predictions
const predictedLabels = encoder.inverseTransform(predictions);
console.log('Predicted original labels:', predictedLabels);
// Output: Predicted original labels: [10, 20, 30, 10]
Example: Error Handling
import { LabelEncoder } from 'bun-scikit';
const encoder = new LabelEncoder();
encoder.fit([1, 2, 3]);
// Try to encode unknown label
try {
encoder.transform([1, 2, 4]); // 4 was not seen during fit
} catch (error) {
console.error(error.message);
// Output: Unknown label 4 at index 2.
}
// Try to decode invalid index
try {
encoder.inverseTransform([0, 1, 5]); // 5 is out of range
} catch (error) {
console.error(error.message);
// Output: Encoded label out of range at index 2: 5.
}
Notes
- Labels are sorted before encoding, so encoding is consistent across fits
- All labels in
transform() must have been seen during fit()
- Use OneHotEncoder if you need one-hot encoded output instead of integer labels
- The encoder must be fitted before calling
transform() or inverseTransform()
- Input must be finite numeric values