StandardScaler

Standard scaler is used to standardize features by removing the mean and scaling to unit variance. The standard score of a sample x is calculated as:

Import

import * as datacook from '@pipcook/datacook';
const { StandardScaler } = datacook.Preprocess;

Constructor

const standardScaler = new StandardScaler();

Parameters

Option in params

Properties

mean<Tensor>

Means of current features

standardVariance<Tensor>

Standard variance of current features

withMean <boolean>

Whether to center the data before scaling

withStd <boolean>

Whether to scale the data to unit variance

nFeatures <number>

Number of features

Methods

fit()

Fit standard scaler

Syntax

async fit(X: Tensor | RecursiveArray<number>): Promise<void> 

Parameters

Example

const standardScaler = new StandardScaler();
await standardScaler.fit([[0, 10], [1, 15], [2, 12]]);

fitTransform()

Fit and get transformed score according to input data

Syntax

async fitTransform(X: Tensor | RecursiveArray<number>): Promise<Tensor> 

Parameters

Returns

Promise<Tensor> Promise of transformed data

Example

const standardScaler = new StandardScaler();
const scores= await standardScaler.fitTransform([[0, 10], [1, 15], [2, 12]]);
scores.print();
/**
 * Tensor
 *   [[-1, -0.9271726],
 *    [0 , 1.0596261 ],
 *    [1 , -0.1324531]]
 * /

transform()

Get transformed score according to input data and fitted scaler.

Syntax

async transform(X: Tensor | RecursiveArray<number>): Promise<Tensor>

Parameters

Returns

Promise<Tensor> Promise of transformed data

Example

const standardScaler = new StandardScaler();
await standardScaler.fit([[0, 10], [1, 15], [2, 12]]);
const scores= await standardScaler.transform([[0, 12], [1, 10]]);
scores.print();
/**
 * Tensor
 *   [[-1, -0.1324531],
 *    [0 , -0.9271726]]
 * /

inverseTransform()

Transform score to original scale.

Syntax

inverseTransform(X: Tensor | RecursiveArray<number>): Promise<Tensor>

Parameters

Returns

Promise<Tensor> Promise of data after inverse transforming

Example

const standardScaler = new StandardScaler();
await standardScaler.fit([[0, 10], [1, 15], [2, 12]]);
const data= await standardScaler.inverseTransform([[1, 1], [0, 0.5]]);
data.print();
/**
 * Tensor
 *   [[2, 14.8499441],
 *    [1, 13.5916386]]
 * /