从已经聚类的数据中获取距离矩阵

Question

我使用了 TSclust 包中的 hclust 来进行凝聚层次聚类。我的问题是，我可以从 hclust 取回差异（距离）矩阵吗？我想要距离的值来排名哪个变量更接近变量组中的单个变量。

例子：如果(x1,x2,x3,x4,x5,x6,x7,x8,x9,x10)是构成距离矩阵的变量，那么我想要的是x3和其余变量（x3x1、x3x2、x3x4、x3x5 等）。我们能做到吗？这是代码和可重现的数据。

数据：

structure(list(x1 = c(186.41, 100.18, 12.3, 14.38, 25.97, 0.06, 
0, 6.17, 244.06, 19.26, 256.18, 255.69, 121.88, 75, 121.45, 11.34, 
34.68, 3.09, 34.3, 26.13, 111.31), x2 = c(327.2, 8.05, 4.23, 
6.7, 3.12, 1.91, 37.03, 39.17, 140.06, 83.72, 263.29, 261.22, 
202.48, 23.27, 2.87, 7.17, 14.48, 3.41, 5.95, 70.56, 91.58), 
    x3 = c(220.18, 126.14, 98.59, 8.56, 0.5, 0.9, 17.45, 191.1, 
    164.64, 224.36, 262.86, 237.75, 254.88, 42.05, 9.12, 0.04, 
    12.22, 0.61, 61.86, 114.08, 78.94), x4 = c(90.74, 26.11, 
    47.86, 10.86, 3.74, 23.69, 61.79, 68.12, 87.92, 171.76, 260.98, 
    266.62, 96.27, 57.15, 78.89, 16.73, 6.59, 49.44, 57.21, 202.2, 
    67.17), x5 = c(134.09, 27.06, 7.44, 4.53, 17, 47.66, 95.96, 
    129.53, 40.23, 157.37, 172.61, 248.56, 160.84, 421.94, 109.93, 
    22.77, 2.11, 49.18, 64.13, 52.61, 180.87), x6 = c(173.17, 
    46.68, 6.54, 3.05, 0.35, 0.12, 5.09, 72.46, 58.19, 112.31, 
    233.77, 215.82, 100.63, 65.84, 2.69, 0.01, 3.63, 12.93, 66.55, 
    28, 61.74), x7 = c(157.22, 141.81, 19.98, 116.18, 16.55, 
    122.3, 62.67, 141.84, 78.3, 227.27, 340.22, 351.38, 147.73, 
    0.3, 56.12, 33.2, 5.51, 54.4, 82.98, 152.66, 218.26), x8 = c(274.08, 
    51.92, 54.86, 15.37, 0.31, 0.05, 36.3, 162.04, 171.78, 181.39, 
    310.73, 261.55, 237.99, 123.99, 1.92, 0.74, 0.23, 18.51, 
    7.68, 65.55, 171.33), x9 = c(262.71, 192.34, 2.75, 21.68, 
    1.69, 3.92, 0.09, 9.33, 120.36, 282.92, 236.7, 161.59, 255.44, 
    126.44, 7.63, 2.04, 1.02, 0.12, 5.87, 146.25, 134.11), x10 = c(82.71, 
    44.09, 1.52, 2.63, 4.38, 28.64, 168.43, 80.62, 20.36, 39.29, 
    302.31, 247.52, 165.73, 18.27, 2.67, 1.77, 23.13, 53.47, 
    53.14, 46.61, 86.29)), class = "data.frame", row.names = c(NA, 
-21L))

Code:

as.ts(cdata)
library(dplyr) # data wrangling
library(ggplot2) # grammar of graphics
library(ggdendro) # dendrograms
library(TSclust) # cluster time series

cluster analysis

dist_ts <- TSclust::diss(SERIES = t(cdata), METHOD = "INT.PER") # note the data frame must be transposed
hc <- stats::hclust(dist_ts, method="complete") # method can be also "average" or diana (for DIvisive ANAlysis Clustering)
hcdata <- ggdendro::dendro_data(hc)
names_order <- hcdata$labels$label
# Use the following to remove labels from dendogram so not doubling up - but good for checking hcdata$labels$label <- ""
hcdata%>%ggdendro::ggdendrogram(., rotate=FALSE, leaf_labels=FALSE)

Answer 1

我相信您要查找的对象存储在变量 dist_ts:

dist_ts <- TSclust::diss(SERIES = t(cdata), METHOD = "INT.PER")
print(dist_ts)

从已经聚类的数据中获取距离矩阵

Getting the distance matrix back from already clustered data

r

cluster-analysis

data-mining

hierarchical-clustering