robust_scaler

A Rust implementation of scikit-learn's RobustScaler, designed for robust feature scaling in machine learning pipelines.

RobustScaler removes the median and scales data according to the interquartile range (IQR), making it robust to outliers.

This crate is ideal for preprocessing input data in ML services, especially when integrating with models trained in Python and deployed in Rust.

🚀 Features

• ✅ Compatible with scikit-learn's RobustScaler
• ✅ Load from JSON (exported from sklearn)
• ✅ 1D and 2D data support
• ✅ No external dependencies beyond ndarray and serde
• ✅ Thread-safe usage with Lazy or Arc
• ✅ Perfect for web APIs (e.g., Actix, Axum)

📦 Installation

Add to your Cargo.toml:

robust_scaler = "0.1.0"

Or use locally during development:

robust_scaler = { path = "./robust_scaler" }

🔧 Usage

Load from JSON (recommended for production)

use robust_scaler::RobustScaler;

let scaler = RobustScaler::from_json("robust_scaler.json")
    .expect("Failed to load scaler");

let scaled = scaler.transform_1d(&[1.0, 2.0, 3.0]);

The JSON should be exported from scikit-learn like this:

import json
from sklearn.preprocessing import RobustScaler

scaler = RobustScaler()
scaler.fit(data)

with open("robust_scaler.json", "w") as f:
    json.dump({
        "center_": scaler.center_.tolist(),
        "scale_": scaler.scale_.tolist(),
        "n_features_in_": scaler.n_features_in_
    }, f)

Fit and transform in Rust (for testing)

use ndarray::arr2;
use robust_scaler::RobustScaler;

let data = arr2(&[[1.0, 2.0], [3.0, 4.0], [5.0, 6.0]]);
let mut scaler = RobustScaler::new();
scaler.fit(&data);

let scaled = scaler.transform(&data);