PCA (Principle Component Analysis)
Principal component analysis (PCA) is the process of computing the principal components and using them to perform a change of basis on the data, sometimes using only the first few principal components and ignoring the rest.
Import
import * as Datacook from 'datacook';
const { PCA } = DataCook.Model;
Constructor
const pca = new PCA({ nComponent: 5, method: 'correlation' });
Option Parameters
| parameter | type | description |
|---|---|---|
| nComponent | number | the estimated number of components. If not specified, it equals the lesser value of nFeatures and nSamples |
| method | ‘covariance’|’correlation’ | Method used for decomposition. default=’covariance’. ‘covariance’: use covariance matrix for decomposition. ‘correlation’: use correlation matrix for decomposition. |
Methods
fit
Fit PCA model with input xData
Syntax
async fit(xData: Tensor | RecursiveArray<number>): Promise<void>
Paramters
| Parameter | type | description |
|---|---|---|
| xData | Tensor | RecursiveArray | input data of shape (nSamples,nFeatures) in type of array or tensor |
transform
ransform method of PCA, transform original data to reduced features.
Syntax
async transform(xData: Tensor | RecursiveArray<number>): Promise<Tensor>
Parameters
| Parameter | type | description |
|---|---|---|
| xData | Tensor | RecursiveArray | input data of shape (nSamples,nFeatures) in type of array or tensor |
Example
PCA of iris data
import * as DataCook from '@pipcook/datacook';
import {PCA} from DataCook.Model;
const pca = new PCA({ nComponents: 3 });
await pca.fit(irisData);
const fittedNTensors = tf.memory().numTensors;
assert.isTrue(fittedNTensors === nTensorsBase + 5);
console.log('explained variance:', pca.explainedVariance.arraySync());
console.log('explained varaince ratio:', pca.explainedVarianceRatio.arraySync());