README.md

In DavidGMarquez/evolvingClusteringR: Evolving Clustering

evolvingClusteringR

A R package for evolving clustering that implements the evolving K-means and evolving EM algorithms described in "A novel and simple strategy for evolving prototype based clustering" Paper

Installation

Use install_github from devtools package

library("devtools")
install_github("DavidGMarquez/evolvingClusteringR")

Usage

There are two functions in the package testEvolutiveDynamicConfluenceFastKmeans and testEvolutiveDynamicConfluenceFastEM. Both functions take as arguments the path of a CSV file, the number of samples for the initialization and a list of parameters. The CSV file should have just numbers separated by commas, the last column must be the ground truth. In the list of parameters you should add the memory parameter of the algorithm (see Paper ) and the number of clusters for the initialization KinitMclust.

In the folder Scripts is also included the file exampleMain.R. This code executes the algorithms over the GaussianMotionData dataset.

Examples

 tf = tempfile()
 iris2 = iris
 iris2$Species = as.numeric(iris2$Species)
 iris2 <- iris2[sample(nrow(iris2)),]
 write.table(iris2,tf,row.names=FALSE, col.names=FALSE,sep=",")
 samplesToInitMeta=50
 parameters=list(memory=2,KinitMclust=3)
 resultsA=testEvolutiveDynamicConfluenceFastEM(tf,samplesToInitMeta,parameters)
 resultsA$clusterModel$fit
 plot(iris2[,-5],col=resultsA$clusterModel$fit)
 parameters<-list(memory=0.01,KinitMclust=3)
 resultsB=testEvolutiveDynamicConfluenceFastEM(tf,samplesToInitMeta,parameters)
 resultsB$clusterModel$fit
 plot(iris2[,-5],col=resultsB$clusterModel$fit)

 tf = tempfile()
 iris2 = iris
 iris2$Species = as.numeric(iris2$Species)
 iris2 <- iris2[sample(nrow(iris2)),]
 write.table(iris2,tf,row.names=FALSE, col.names=FALSE,sep=",")
 samplesToInitMeta=50
 parameters=list(memory=2,KinitMclust=3)
 resultsA=testEvolutiveDynamicConfluenceFastKmeans(tf,samplesToInitMeta,parameters)
 resultsA$clusterModel$fit
 plot(iris2[,-5],col=resultsA$clusterModel$fit)
 parameters<-list(memory=0.01,KinitMclust=3)
 resultsB=testEvolutiveDynamicConfluenceFastKmeans(tf,samplesToInitMeta,parameters)
 resultsB$clusterModel$fit
 plot(iris2[,-5],col=resultsB$clusterModel$fit)

Evolving EM Results Animation

Reference

Please cite this publication when referencing or using this material:

@article{MARQUEZ201816,
title = "A novel and simple strategy for evolving prototype based clustering",
journal = "Pattern Recognition",
volume = "82",
pages = "16 - 30",
year = "2018",
issn = "0031-3203",
doi = "https://doi.org/10.1016/j.patcog.2018.04.020",
url = "http://www.sciencedirect.com/science/article/pii/S0031320318301547",
author = "David G. Márquez and Abraham Otero and Paulo Félix and Constantino A. García",
keywords = "Evolving clustering, Data stream, Concept drift, Gaussian mixture models, K-means, Cluster evolution"
}

License

The package is available under GNU General Public License v3.0 gpl-3.0. We encourage other researchers to use it and send their feeback.

Maintained by David Gonzalez Marquez.

DavidGMarquez/evolvingClusteringR documentation built on May 22, 2019, 2:01 p.m.