1

对于此数据如何解决此问题

> x=data.frame(c(v1="a" ,"b" ,"c" ,"d" ,"e"),
+ v2=c(97 ,90 ,93 ,97 ,90),
+ v3=c( 85 ,91 ,87 ,91 ,93))
> library(e1071)
> f <- cmeans(x, 2)
Error in cmeans(x, 2) : NA/NaN/Inf in foreign function call (arg 1)
In addition: Warning messages:
1: In cmeans(x, 2) : NAs introduced by coercion
2: In cmeans(x, 2) : NAs introduced by coercion
> f

我想将 c-maen 应用于我的数据，如上面的代码所示，它包含三个向量：v1、v2、v2 我想通过向量（v1）应用 c-mean 标签

4

2 回答 2

2

如果我们查看 的文档?cmeans，

x - 列对应变量和行对应观察的数据矩阵。

因此，我们可以在删除字符列（第一列）后转换data.frame为matrix

x1 <- as.matrix(x[-1])
row.names(x1) <- x[,1]
cmeans(x1, 2)
#Fuzzy c-means clustering with 2 clusters

#Cluster centers:
#        v2       v3
#1 90.30090 91.85191
#2 95.75436 87.22535

#Memberships:
#           1          2
#a 0.06614213 0.93385787
#b 0.98305641 0.01694359
#c 0.19855988 0.80144012
#d 0.25730888 0.74269112
#e 0.97924422 0.02075578

#Closest hard clustering:
#a b c d e 
#2 1 2 2 1 

#Available components:
#[1] "centers"     "size"        "cluster"     "membership"  "iter"        "withinerror" "call"
于 2019-05-09T21:06:02.830 回答
0

分区聚类算法家族的k-mean工作原理是mean其本质上只接受数值。您收到错误，因为数据框由数字和分类值组成，c-mean()不喜欢。此外，无需将数据帧转换为矩阵，因为这不是实际问题。

所以，

替代方法

离散字符变量以分配数字，然后应用聚类。这样就不需要删除任何变量。

# create empty data frame
df<- setNames(data.frame(matrix(ncol = 5, nrow = 5)), c("a" ,"b" ,"c" ,"d" ,"e"))

# fill values
df$a<- c("aaaa" ,"bbbb" ,"cccc" ,"dddd" ,"eeee")
df$b<- c(97 ,90 ,93 ,97 ,90)
df$c<- c(97 ,90 ,93 ,97 ,90)
df$d<- c( 85 ,91 ,87 ,91 ,93)
df$e<- c( 85 ,91 ,87 ,91 ,93)

# show the dataframe
df
 a  b  c  d  e
1 aaaa 97 97 85 85
2 bbbb 90 90 91 91
3 cccc 93 93 87 87
4 dddd 97 97 91 91
5 eeee 90 90 93 93

# Discretize the character variable
df$a <- as.numeric( factor(df$a) ) -1
df
  a  b  c  d  e
1 0 97 97 85 85
2 1 90 90 91 91
3 2 93 93 87 87
4 3 97 97 91 91
5 4 90 90 93 93

# Apply clustering
library(e1071)
cmeans(df, 2)
Fuzzy c-means clustering with 2 clusters

Cluster centers:
      a     b     c     d     e
1 1.406 95.72 95.72 87.18 87.18
2 2.510 90.36 90.36 91.85 91.85

Memberships:
           1       2
[1,] 0.92728 0.07272
[2,] 0.04014 0.95986
[3,] 0.80061 0.19939
[4,] 0.72009 0.27991
[5,] 0.03544 0.96456

Closest hard clustering:
[1] 1 2 1 1 2

Available components:
[1] "centers"     "size"        "cluster"     "membership"  "iter"       
[6] "withinerror" "call"
于 2019-05-10T06:17:50.017 回答