Package org.hipparchus.stat.projection

Class PCA

java.lang.Object

org.hipparchus.stat.projection.PCA

public class PCA
extends Object

Principal component analysis (PCA) is a statistical technique for reducing the dimensionality of a dataset. PCA can be thought of as a projection or scaling of the data to reduce the number of dimensions but done in a way that preserves as much information as possible.

Since:

3.0

Constructor Summary

Constructors

Constructor	Description
PCA(int numC)	A default PCA will center but not scale.
PCA(int numC, boolean scale, boolean biasCorrection)	Create a PCA with the ability to adjust scaling parameters.

Method Summary

Modifier and Type	Method	Description
PCA	fit(double[][] data)	Fit our model to the data, ready for subsequence transforms.
double[][]	fitAndTransform(double[][] data)	Fit our model to the data and then transform it to the reduced dimensions.
double[]	getCenter()	Get by column center (or mean) of the fitted data.
double[][]	getComponents()	Returns the principal components of our projection model.
int	getNumComponents()	GEt number of components.
double[]	getVariance()	Get principal component variances.
boolean	isBiasCorrection()	Check whether scaling (correlation), if in use, adjusts for bias.
boolean	isScale()	Check whether scaling (correlation) or no scaling (covariance) is used.
double[][]	transform(double[][] data)	Transform the supplied data using our projection model.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

PCA

public PCA(int numC,
           boolean scale,
           boolean biasCorrection)

Create a PCA with the ability to adjust scaling parameters.

Parameters:

numC - the number of components

scale - whether to also scale (correlation) rather than just center (covariance)

biasCorrection - whether to adjust for bias when scaling

PCA

public PCA(int numC)

A default PCA will center but not scale.

Parameters:

numC - the number of components

Method Detail

getNumComponents

public int getNumComponents()

GEt number of components.

Returns:

the number of components

isScale

public boolean isScale()

Check whether scaling (correlation) or no scaling (covariance) is used.

Returns:

whether scaling (correlation) or no scaling (covariance) is used

isBiasCorrection

public boolean isBiasCorrection()

Check whether scaling (correlation), if in use, adjusts for bias.

Returns:

whether scaling (correlation), if in use, adjusts for bias

getVariance

public double[] getVariance()

Get principal component variances.

Returns:

the principal component variances, ordered from largest to smallest, which are the eigenvalues of the covariance or correlation matrix of the fitted data

getCenter

public double[] getCenter()

Get by column center (or mean) of the fitted data.

Returns:

the by column center (or mean) of the fitted data

getComponents

public double[][] getComponents()

Returns the principal components of our projection model. These are the eigenvectors of our covariance/correlation matrix.

Returns:

the principal components

fitAndTransform

public double[][] fitAndTransform(double[][] data)

Fit our model to the data and then transform it to the reduced dimensions.

Parameters:

data - the input data

Returns:

the fitted data

transform

public double[][] transform(double[][] data)

Transform the supplied data using our projection model.

Parameters:

data - the input data

Returns:

the fitted data

fit

public PCA fit(double[][] data)

Fit our model to the data, ready for subsequence transforms.

Parameters:

data - the input data

Returns:

this