smile.feature

Class SumSquaresRatio

java.lang.Object

smile.feature.SumSquaresRatio

All Implemented Interfaces:

FeatureRanking

public class SumSquaresRatio
extends java.lang.Object
implements FeatureRanking

The ratio of between-groups to within-groups sum of squares is a univariate feature ranking metric, which can be used as a feature selection criterion for multi-class classification problems. For each variable j, this ratio is BSS(j) / WSS(j) = ΣI(y_i = k)(x_kj - x_·j)² / ΣI(y_i = k)(x_ij - x_kj)²; where x_·j denotes the average of variable j across all samples, x_kj denotes the average of variable j across samples belonging to class k, and x_ij is the value of variable j of sample i. Clearly, features with larger sum squares ratios are better for classification.

References

1. S. Dudoit, J. Fridlyand and T. Speed. Comparison of discrimination methods for the classification of tumors using gene expression data. J Am Stat Assoc, 97:77-87, 2002.

Constructor Summary

Constructors

Constructor and Description
SumSquaresRatio()

Method Summary

Modifier and Type	Method and Description
double[]	rank(double[][] x, int[] y) Univariate feature ranking.

Methods inherited from class java.lang.Object

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

Constructor Detail

SumSquaresRatio

public SumSquaresRatio()

Method Detail

rank

public double[] rank(double[][] x,
                     int[] y)

Description copied from interface: FeatureRanking

Univariate feature ranking. Note that this method actually does NOT rank the features. It just returns the metric values of each feature. The caller can then rank and select features.

Specified by:

rank in interface FeatureRanking

Parameters:

x - a n-by-p matrix of n instances with p features.

y - class labels in [0, k), where k is the number of classes.

Returns:

the metric values of each feature.