starlight-preprocess

Starlight Preprocess is a lightweight preprocessing library for machine learning in the Starlight ecosystem. It provides common data transformation tools such as scaling, encoding, and dataset splitting — designed to work seamlessly with other Starlight ML packages.

Features

• Feature Scaling

• StandardScaler (mean = 0, std = 1)

• MinMaxScaler (range 0–1)

• Label Encoding

• LabelEncoder (class → integer)

• OneHotEncoder (categorical → binary vectors)

• Dataset Splitting

• Train / Test split utility

• Pipeline Friendly

• Works with starlight-regression, starlight-classifier, starlight-cluster

• Designed for starlight-pipeline

Installation

npm install starlight-preprocess

Usage Examples

Standard Scaling

import { StandardScaler } from "starlight-preprocess";

const X = [
  [1, 10],
  [2, 20],
  [3, 30]
];

const scaler = new StandardScaler();
const XScaled = scaler.fitTransform(X);

console.log(XScaled);

Min-Max Scaling

import { MinMaxScaler } from "starlight-preprocess";

const scaler = new MinMaxScaler();
const normalized = scaler.fitTransform(X);

Label Encoding

import { LabelEncoder } from "starlight-preprocess";

const labels = ["cat", "dog", "cat", "bird"];

const encoder = new LabelEncoder();
const encoded = encoder.fitTransform(labels);

console.log(encoded);
// [0, 1, 0, 2]

One-Hot Encoding

import { OneHotEncoder } from "starlight-preprocess";

const encoder = new OneHotEncoder();
const oneHot = encoder.fitTransform(["red", "blue", "red"]);

console.log(oneHot);
// [[1,0],[0,1],[1,0]]

Train / Test Split

import { trainTestSplit } from "starlight-preprocess";

const X = [[1], [2], [3], [4], [5]];
const y = [2, 4, 6, 8, 10];

const { XTrain, XTest, yTrain, yTest } =
  trainTestSplit(X, y, 0.2);

console.log(XTrain, XTest);

API Overview

Scalers

• StandardScaler
• MinMaxScaler

Methods

• fit(X)
• transform(X)
• fitTransform(X)

Encoders

• LabelEncoder
• OneHotEncoder

Methods

• fit(values)
• transform(values)
• fitTransform(values)
• inverseTransform() (LabelEncoder)

Utilities

• trainTestSplit(X, y, testSize = 0.2, shuffle = true)

Works Great With

• starlight-ml
• starlight-vec
• starlight-classifier
• starlight-regression
• starlight-cluster
• starlight-pipeline

License

MIT License © Dominex Macedon

Roadmap

• Missing value handling
• Text normalization helpers
• Image preprocessing
• Auto-preprocessing in pipelines