探索性数据分析week3

3.1 Hierarchical Clustering层次聚类

method可为"ward.D", "ward.D2", "single", "complete", "average" (= UPGMA), "mcquitty" (= WPGMA), "median" (= WPGMC) or "centroid" (= UPGMC).

也就是常说的全链或那什么层次聚类

Can we find things that are close together?

Clustering organizes things that are close into groups

• How do we define close?
• How do we group things?
• How do we visualize the grouping?
• How do we interpret the grouping?

http://scholar.google.com/scholar?hl=en&q=cluster+analysis&btnG=&as_sdt=1%2C21&as_sdtp=

一个关于聚类论文的链接，除非翻墙，否则是打不开的了

Hierarchical clustering

• An agglomerative approach（一种凝聚的方法）
  • Find closest two things
  • Put them together
  • Find next closest
• Requires
  • A defined distance
  • A merging approach
• Produces
  • A tree showing how close things are to each other

How do we define close?

• Most important step
  • Garbage in -> garbage out
• Distance or similarity
  • Continuous - euclidean distance
  • Continuous - correlation similarity
  • Binary - manhattan distance
• Pick a distance/similarity that makes sense for your problem

http://rafalab.jhsph.edu/688/lec/lecture5-clustering.pdf

Hierarchical clustering - dist

• Important parameters: x,method

dataFrame <- data.frame(x = x, y = y)
dist(dataFrame)

##          1       2       3       4       5       6       7       8       9
## 2  0.34121                                                                
## 3  0.57494 0.24103                                                        
## 4  0.26382 0.52579 0.71862                                                
## 5  1.69425 1.35818 1.11953 1.80667                                        
## 6  1.65813 1.31960 1.08339 1.78081 0.08150                                
## 7  1.49823 1.16621 0.92569 1.60132 0.21110 0.21667                        
## 8  1.99149 1.69093 1.45649 2.02849 0.61704 0.69792 0.65063                
## 9  2.13630 1.83168 1.67836 2.35676 1.18350 1.11500 1.28583 1.76461        
## 10 2.06420 1.76999 1.63110 2.29239 1.23848 1.16550 1.32063 1.83518 0.14090
## 11 2.14702 1.85183 1.71074 2.37462 1.28154 1.21077 1.37370 1.86999 0.11624
## 12 2.05664 1.74663 1.58659 2.27232 1.07701 1.00777 1.17740 1.66224 0.10849

下面还有很多，此处略下不表

 
 
 

  
  
  ## Hierarchical clustering - #1
 
 
 
 
 
 

  
  
  

 
 
 
 
 
 

  
  
  ```{r dependson="createData",echo=FALSE, fig.height=4,fig.width=8}
 
 
 
 
 
 

  
  
  suppressMessages(library(fields))
 
 
 
 
 
 

  
  
  dataFrame <- data.frame(x=x,y=y)
 
 
 
 
 
 

  
  
  rdistxy <- rdist(dataFrame)
 
 
 
 
 
 

  
  
  diag(rdistxy) <- diag(rdistxy) + 1e5
 
 
 
 
 
 

  
  
  

 
 
 
 
 
 

  
  
  # Find the index of the points with minimum distance
 
 
 
 
 
 

  
  
  ind <- which(rdistxy == min(rdistxy),arr.ind=TRUE)
 
 
 
 
 
 

  
  
  par(mfrow=c(1,2),mar=rep(0.2,4))
 
 
 
 
 
 

  
  
  # Plot the points with the minimum overlayed
 
 
 
 
 
 

  
  
  plot(x,y,col="blue",pch=19,cex=2)
 
 
 
 
 
 

  
  
  text(x+0.05,y+0.05,labels=as.character(1:12))
 
 
 
 
 
 

  
  
  points(x[ind[1,]],y[ind[1,]],col="orange",pch=19,cex=2)
 
 
 
 
 
 

  
  
  

 
 
 
 
 
 

  
  
  # Make a cluster and cut it at the right height
 
 
 
 
 
 

  
  
  distxy <- dist(dataFrame)
 
 
 
 
 
 

  
  
  hcluster <- hclust(distxy)
 
 
 
 
 
 

  
  
  dendro <- as.dendrogram(hcluster)
 
 
 
 
 
 

  
  
  cutDendro <- cut(dendro,h=(hcluster$height[1]+0.00001) )
 
 
 
 
 
 

  
  
  plot(cutDendro$lower[[11]],yaxt="n")
 
 
 
 
 
 

  
  
  ```
 
 
 

 
 
 

  
  
  ## Hierarchical clustering - #2
 
 
 
 
 
 

  
  
  

 
 
 
 
 
 

  
  
  ```{r dependson="createData",echo=FALSE}
 
 
 
 
 
 

  
  
  library(fields)
 
 
 
 
 
 

  
  
  dataFrame <- data.frame(x=x,y=y)
 
 
 
 
 
 

  
  
  rdistxy <- rdist(dataFrame)
 
 
 
 
 
 

  
  
  diag(rdistxy) <- diag(rdistxy) + 1e5
 
 
 
 
 
 

  
  
  

 
 
 
 
 
 

  
  
  # Find the index of the points with minimum distance
 
 
 
 
 
 

  
  
  ind <- which(rdistxy == min(rdistxy),arr.ind=TRUE)
 
 
 
 
 
 

  
  
  par(mar=rep(0.2,4))
 
 
 
 
 
 

  
  
  # Plot the points with the minimum overlayed
 
 
 
 
 
 

  
  
  plot(x,y,col="blue",pch=19,cex=2)
 
 
 
 
 
 

  
  
  text(x+0.05,y+0.05,labels=as.character(1:12))
 
 
 
 
 
 

  
  
  points(x[ind[1,]],y[ind[1,]],col="orange",pch=19,cex=2)
 
 
 
 
 
 

  
  
  points(mean(x[ind[1,]]),mean(y[ind[1,]]),col="black",cex=3,lwd=3,pch=3)
 
 
 
 
 
 

  
  
  points(mean(x[ind[1,]]),mean(y[ind[1,]]),col="orange",cex=5,lwd=3,pch=1)
 
 
 

原来这里的大圆不过就是把pch=1变大而已，日无语啊，还能再弱智点不？

 
 
 

  
  
  # Find the index of the points with minimum distance
 
 
 
 
 
 

  
  
  ind <- which(rdistxy == rdistxy[order(rdistxy)][3],arr.ind=TRUE)
 
 
 
 
 
 

  
  
  par(mfrow=c(1,3),mar=rep(0.2,4))
 
 
 
 
 
 

  
  
  # Plot the points with the minimum overlayed
 
 
 
 
 
 

  
  
  plot(x,y,col="blue",pch=19,cex=2)
 
 
 
 
 
 

  
  
  text(x+0.05,y+0.05,labels=as.character(1:12))
 
 
 
 
 
 

  
  
  points(x[c(5,6)],y[c(5,6)],col="orange",pch=19,cex=2)
 
 
 
 
 
 

  
  
  points(x[ind[1,]],y[ind[1,]],col="red",pch=19,cex=2)
 
 
 
 
 
 

  
  
  

 
 
 
 
 
 

  
  
  # Make dendogram plots
 
 
 
 
 
 

  
  
  distxy <- dist(dataFrame)
 
 
 
 
 
 

  
  
  hcluster <- hclust(distxy)
 
 
 
 
 
 

  
  
  dendro <- as.dendrogram(hcluster)
 
 
 
 
 
 

  
  
  cutDendro <- cut(dendro,h=(hcluster$height[2]) )
 
 
 
 
 
 

  
  
  plot(cutDendro$lower[[10]],yaxt="n")
 
 
 
 
 
 

  
  
  plot(cutDendro$lower[[5]],yaxt="n")
 
 
 

此处对ind的取值很值得学习啊

Prettier dendrograms（树形图）

myplclust <- function(hclust, lab = hclust$labels, lab.col = rep(1, length(hclust$labels)), 
    hang = 0.1, ...) {
   
    ## modifiction of plclust for plotting hclust objects *in colour*!  Copyright
    ## Eva KF Chan 2009 Arguments: hclust: hclust object lab: a character vector
    ## of labels of the leaves of the tree lab.col: colour for the labels;
    ## NA=default device foreground colour hang: as in hclust & plclust Side
    ## effect: A display of hierarchical cluster with coloured leaf labels.
    y <- rep(hclust$height, 2)
    x <- as.numeric(hclust$merge)
    y