Scribd Logo
Fazer o upload

Bem-vindo(a) ao Scribd!

  • K Means Clustering

    Enviado por

    sasha_232
    52 visualizações14 páginas
    This document discusses the k-means clustering algorithm. It explains that k-means clustering groups data points into k number of clusters defined by centroids, where each data point belongs to the cluster with the nearest centroid. It requires defining a distance metric and providing an initial set of centroids. The algorithm then iteratively reassigns points to centroids and recalculates the centroids until clustering stabilization. Examples are provided using R code to demonstrate k-means clustering on simulated data.

    Descrição original:

    Título original

    k Means Clustering

    Direitos autorais

    © © All Rights Reserved

    Formatos disponíveis

    PDF, TXT ou leia online no Scribd

    Você considera este documento útil?

    Este conteúdo é inapropriado?

    Denunciar este documento
    This document discusses the k-means clustering algorithm. It explains that k-means clustering groups data points into k number of clusters defined by centroids, where each data point belongs to the cluster with the nearest centroid. It requires defining a distance metric and providing an initial set of centroids. The algorithm then iteratively reassigns points to centroids and recalculates the centroids until clustering stabilization. Examples are provided using R code to demonstrate k-means clustering on simulated data.

    Direitos autorais:

    © All Rights Reserved
    Baixe no formato PDF, TXT ou leia online no Scribd
    Sinalizar o conteúdo como inadequado
    52 visualizações14 páginas

    K Means Clustering

    Enviado por

    sasha_232
    This document discusses the k-means clustering algorithm. It explains that k-means clustering groups data points into k number of clusters defined by centroids, where each data point belongs to the cluster with the nearest centroid. It requires defining a distance metric and providing an initial set of centroids. The algorithm then iteratively reassigns points to centroids and recalculates the centroids until clustering stabilization. Examples are provided using R code to demonstrate k-means clustering on simulated data.

    Direitos autorais:

    © All Rights Reserved
    Baixe no formato PDF, TXT ou leia online no Scribd
    Sinalizar o conteúdo como inadequado
    de 14

    2/3/13 8:06 PM K-means clustering

    Page 1 of 14 file:///Users/jtleek/Dropbox/Jeff/teaching/2013/coursera/week3/005kmeansClustering/index.html#1
    K-means clustering K-means clustering
    Jeffrey Leek, Assistant Professor of Biostatistics
    Johns Hopkins Bloomberg School of Public Health
    2/3/13 8:06 PM K-means clustering
    Page 2 of 14 file:///Users/jtleek/Dropbox/Jeff/teaching/2013/coursera/week3/005kmeansClustering/index.html#1
    Can we find things that are close together? Can we find things that are close together?
    How do we define close?
    How do we group things?
    How do we visualize the grouping?
    How do we interpret the grouping?

    2/14
    2/3/13 8:06 PM K-means clustering
    Page 3 of 14 file:///Users/jtleek/Dropbox/Jeff/teaching/2013/coursera/week3/005kmeansClustering/index.html#1
    How do we define close? How do we define close?
    Most important step
    Distance or similarity
    Pick a distance/similarity that makes sense for your problem

    Garbage in -> garbage out -

    Continuous - euclidean distance


    Continous - correlation similarity
    Binary - manhattan distance
    -
    -
    -

    3/14
    2/3/13 8:06 PM K-means clustering
    Page 4 of 14 file:///Users/jtleek/Dropbox/Jeff/teaching/2013/coursera/week3/005kmeansClustering/index.html#1
    K-means clustering K-means clustering
    A partioning approach
    Requires
    Produces

    Fix a number of clusters


    Get "centroids" of each cluster
    Assign things to closest centroid
    Reclaculate centroids
    -
    -
    -
    -

    A defined distance metric


    A number of clusters
    An initial guess as to cluster centroids
    -
    -
    -

    Final estimate of cluster centroids


    An assignment of each point to clusters
    -
    -
    4/14
    2/3/13 8:06 PM K-means clustering
    Page 5 of 14 file:///Users/jtleek/Dropbox/Jeff/teaching/2013/coursera/week3/005kmeansClustering/index.html#1
    K-means clustering - K-means clustering - example example
    set.seed(1234); par(mar=c(0,0,0,0))
    x <- rnorm(12,mean=rep(1:3,each=4),sd=0.2)
    y <- rnorm(12,mean=rep(c(1,2,1),each=4),sd=0.2)
    plot(x,y,col="blue",pch=19,cex=2)
    text(x+0.05,y+0.05,labels=as.character(1:12))
    5/14
    2/3/13 8:06 PM K-means clustering
    Page 6 of 14 file:///Users/jtleek/Dropbox/Jeff/teaching/2013/coursera/week3/005kmeansClustering/index.html#1
    K-means clustering - K-means clustering - starting centroids starting centroids
    6/14
    2/3/13 8:06 PM K-means clustering
    Page 7 of 14 file:///Users/jtleek/Dropbox/Jeff/teaching/2013/coursera/week3/005kmeansClustering/index.html#1
    K-means clustering - K-means clustering - assign to closest centroid assign to closest centroid
    7/14
    2/3/13 8:06 PM K-means clustering
    Page 8 of 14 file:///Users/jtleek/Dropbox/Jeff/teaching/2013/coursera/week3/005kmeansClustering/index.html#1
    K-means clustering - K-means clustering - recalculate centroids recalculate centroids
    8/14
    2/3/13 8:06 PM K-means clustering
    Page 9 of 14 file:///Users/jtleek/Dropbox/Jeff/teaching/2013/coursera/week3/005kmeansClustering/index.html#1
    K-means clustering - K-means clustering - reassign values reassign values
    9/14
    2/3/13 8:06 PM K-means clustering
    Page 10 of 14 file:///Users/jtleek/Dropbox/Jeff/teaching/2013/coursera/week3/005kmeansClustering/index.html#1
    K-means clustering - K-means clustering - update centroids update centroids
    10/14
    2/3/13 8:06 PM K-means clustering
    Page 11 of 14 file:///Users/jtleek/Dropbox/Jeff/teaching/2013/coursera/week3/005kmeansClustering/index.html#1
    kmeans() kmeans()
    Important parameters: x,centers,iter.max,nstart
    dataFrame <- data.frame(x,y)
    kmeansObj <- kmeans(dataFrame,centers=3)
    names(kmeansObj)
    [1] "cluster" "centers" "totss" "withinss" "tot.withinss" "betweenss"
    [7] "size"
    kmeansObj$cluster
    [1] 3 3 3 3 1 1 1 1 2 2 2 2
    11/14
    2/3/13 8:06 PM K-means clustering
    Page 12 of 14 file:///Users/jtleek/Dropbox/Jeff/teaching/2013/coursera/week3/005kmeansClustering/index.html#1
    kmeans() kmeans()
    par(mar=rep(0.2,4))
    plot(x,y,col=kmeansObj$cluster,pch=19,cex=2)
    points(kmeansObj$centers,col=1:3,pch=3,cex=3,lwd=3)
    12/14
    2/3/13 8:06 PM K-means clustering
    Page 13 of 14 file:///Users/jtleek/Dropbox/Jeff/teaching/2013/coursera/week3/005kmeansClustering/index.html#1
    Heatmaps Heatmaps
    set.seed(1234)
    dataMatrix <- as.matrix(dataFrame)[sample(1:12),]
    kmeansObj2 <- kmeans(dataMatrix,centers=3)
    par(mfrow=c(1,2),mar=rep(0.2,4))
    image(t(dataMatrix)[,nrow(dataMatrix):1],yaxt="n")
    image(t(dataMatrix)[,order(kmeansObj$cluster)],yaxt="n")
    13/14
    2/3/13 8:06 PM K-means clustering
    Page 14 of 14 file:///Users/jtleek/Dropbox/Jeff/teaching/2013/coursera/week3/005kmeansClustering/index.html#1
    Notes and further resources Notes and further resources
    K-means requires a number of clusters
    K-means is not deterministic
    Rafa's Distances and Clustering Video
    Elements of statistical learning

    Pick by eye/intuition
    Pick by cross validation/information theory, etc.
    Determining the number of clusters
    -
    -
    -

    Different # of clusters
    Different number of iterations
    -
    -

    14/14

    Você também pode gostar