PRTools Contents

PRTools User Guide

mlrc

MLRC

MLRC Muli-response Linear Regression Combiner

    W = A*(WU*MLRC)
    W = WT*MLRC(B*WT)
    D = C*W

Input
 A Dataset used for training base classifiers as well as combiner
 B Dataset used for training combiner of trained base classifiers
 C Dataset used for testing (executing) the combiner
 WU Set of untrained base classifiers, see STACKED
 WT Set of trained base classifiers, see STACKED

Output
 W Trained Muli-response Linear Regression Combiner
 D Dataset with prob. products (over base classifiers) per class

Description

Using dataset A that contains the posterior probabilities of each instance  belonging to each class predicted by the base classifiers to train a  multi-response linear regression combiner.  If the original classification problem has K classes, it is converted  into K seperate regression problems, where the problem for class c has  instances with responses equal to 1 when they have label c and zero  otherwise. Put in another way, this function establish a multi-response  linear regression model for each class and utilize these models to estimate  the probability that the instances belong to each class.  Note that in the model for class c, only the probabilities of class c  predicted by the set of base classifiers are used.

Reference(s)

1. Ting, KM, Witten IH. Issues in stacked generalisation, Journal of Artificial Intelligent Research, 1999, 10: 271-289.
2. Dzeroski S, Zenko B. Is combining classifiers with stacking better than selecting the best one? Machine Learning, 2004, 54(3): 255-273.

See also

datasets, mappings, stacked, classc, testd, labeld,

PRTools Contents

PRTools User Guide

This file has been automatically generated. If badly readable, use the help-command in Matlab.