RangeScaler#

RangeScaler implements range-based feature scaling, commonly known as Min-Max scaling.

It scales numeric values in specified columns into a normalized range, typically between 0 and 1, using statistics learned during a fitting phase.

RangeScaler is designed to work with Rowan tables and follows a stateful workflow: statistics are learned during Fit and later applied during Transform.

Overview#

RangeScaler computes the minimum and maximum value for each selected feature and applies the following transformation formula:

scaled = (x - min) / (max - min)

The scaler stores per-column statistics internally and applies transformations in a non-mutating manner, meaning the original table is never modified.

Only numeric columns are considered during fitting and transformation.

Lifecycle#

RangeScaler follows a simple lifecycle:

Create a new scaler instance using NewRangeScaler()
Call Fit() to learn per-column minimum and maximum values
Call Transform() to scale data using the learned statistics
Optionally call Reset() to clear the internal state

Internal State#

RangeScaler maintains the following internal state:

features
A slice of column names that were fitted.
min
A map storing the minimum value for each fitted column.
max
A map storing the maximum value for each fitted column.

This state is updated only during Fit and cleared when Reset() is called.

Constructor#

`NewRangeScaler()`#

NewRangeScaler() creates and initializes a new RangeScaler instance.

The returned scaler starts with an empty internal state and must be fitted before it can be used to transform data.

Example#

import (
	"github.com/go-rowan/rowan"
	"github.com/go-rowan/rowan/scale"
)

func main() {
	tbl, _ := rowan.FromExcel("data.xlsx")

	rangeScaler := scale.NewRangeScaler()
}

Available Methods#

Common Methods#

(*RangeScaler).Fit(t *Table, columns ...string)
(*RangeScaler).Transform(t *Table, columns ...string)
(*RangeScaler).Features()
(*RangeScaler).IsFitted()
(*RangeScaler).Reset()

RangeScaler-Specific Methods#

(*RangeScaler).Min(column string)
(*RangeScaler).Max(column string)

When to Use#

Use RangeScaler when you need to:

Normalize numeric features into a fixed range
Prepare data for algorithms sensitive to feature scale
Apply consistent scaling learned from a reference dataset
Perform min-max normalization in a reproducible way

RangeScaler is especially useful when feature bounds are known or when preserving relative distances within a fixed range is important.