> ## Documentation Index
> Fetch the complete documentation index at: https://docs.clickml.app/llms.txt
> Use this file to discover all available pages before exploring further.

# Feature Scaling component

> Normalize numerical features with standard, min-max, robust, and max-abs scalers to bring columns to a common range before model training.

The **Scaling** component brings all selected numerical features onto the same scale, which improves the performance of distance-based and gradient-based models.

## Configuration

| Option       | Description                                   |
| ------------ | --------------------------------------------- |
| **Method**   | Scaling algorithm to apply (see table below). |
| **Features** | Numerical columns to scale.                   |

### Methods

| Method           | How it works                                                  | Best for                                           |
| ---------------- | ------------------------------------------------------------- | -------------------------------------------------- |
| Standard Scaling | Subtracts mean, divides by std dev. Output has mean=0, std=1. | Most models — SVM, Logistic Regression, KNN        |
| MinMax Scaling   | Rescales values to the `[0, 1]` range.                        | Neural networks, algorithms sensitive to magnitude |
| Robust Scaling   | Uses median and IQR instead of mean and std dev.              | Data with significant outliers                     |

## Input / Output

|        | Type      |
| ------ | --------- |
| Input  | DataFrame |
| Output | DataFrame |

<Note>
  Tree-based models (Random Forest, Decision Tree, Gradient Boosting) do not require scaling. Apply scaling before SVM, KNN, Logistic Regression, and Linear Regression.
</Note>