背景介绍

如果有一组方剂，如下图所示，你想从中挖掘出常用的单味、单味、两味、三味中药，你该怎么去做？

今天我们使用R语言的办法来进行挖掘分析，看看能挖掘出什么东西来？

软件介绍

[软件名称]：R；RStudio

教程介绍

1.首先我们使用在RStudio中设定运行的文件夹，然后加载需要的R包，最后读取数据

# 设定文件夹
setwd("D:Desktop")

# 加载需要的包
# 数据处理绘图用
library(tidyverse)
# 读取excel用
library(openxlsx)
# 关联规则用
library(arules)

# 读取excel数据
data <- read.xlsx("example.xlsx")

2.接下来我们自建个函数，将方剂列分开成单味中药（我构建的这个函数，可以分单个方剂中药物为24个的，自己的是多少可以酌情增加或减少；中药之间使用的是顿号，所以以“、”分隔）

# 自建函数
TCM2dataframe <-  function(TCM){
  # TCM为中药方剂的列，之间使用"、"分割
  TCM_split = TCM%>%str_split("、")
  TCM_rbind = cbind(
    TCM1 =  sapply(TCM_split, "[",1)%>%as.data.frame(),
    TCM2 =  sapply(TCM_split, "[",2)%>%as.data.frame(),
    TCM3 =  sapply(TCM_split, "[",3)%>%as.data.frame(),
    TCM4 =  sapply(TCM_split, "[",4)%>%as.data.frame(),
    TCM5 =  sapply(TCM_split, "[",5)%>%as.data.frame(),
    TCM6 =  sapply(TCM_split, "[",6)%>%as.data.frame(),
    TCM7 =  sapply(TCM_split, "[",7)%>%as.data.frame(),
    TCM8 =  sapply(TCM_split, "[",8)%>%as.data.frame(),
    TCM9 =  sapply(TCM_split, "[",9)%>%as.data.frame(),
    TCM10 = sapply(TCM_split, "[",10)%>%as.data.frame(),
    TCM11 = sapply(TCM_split, "[",11)%>%as.data.frame(),
    TCM12 = sapply(TCM_split, "[",12)%>%as.data.frame(),
    TCM13 = sapply(TCM_split, "[",13)%>%as.data.frame(),
    TCM14 = sapply(TCM_split, "[",14)%>%as.data.frame(),
    TCM15 = sapply(TCM_split, "[",15)%>%as.data.frame(),
    TCM16 = sapply(TCM_split, "[",16)%>%as.data.frame(),
    TCM17 = sapply(TCM_split, "[",17)%>%as.data.frame(),
    TCM18 = sapply(TCM_split, "[",18)%>%as.data.frame(),
    TCM19 = sapply(TCM_split, "[",19)%>%as.data.frame(),
    TCM20 = sapply(TCM_split, "[",20)%>%as.data.frame(),
    TCM21 = sapply(TCM_split, "[",21)%>%as.data.frame(),
    TCM22 = sapply(TCM_split, "[",22)%>%as.data.frame(),
    TCM23 = sapply(TCM_split, "[",23)%>%as.data.frame(),
    TCM24 = sapply(TCM_split, "[",24)%>%as.data.frame())
  
  colnames(TCM_rbind) <- paste0("中药",1:24)
  return(TCM_rbind)
}

# 整合表格并写入csv文件
TCM2dataframe(data$方剂)%>%
  write.csv(file = "TCM_2.csv")

3.读取刚才保存的那个csv文件，查看数据类型前5个

# 读取csv文件
transdata0 <- read.transactions("TCM_2.csv",
                  format = c("basket"),
                  header = TRUE,
                  sep = ",",
                  cols = 1,
                  rm.duplicates = TRUE)

# 查看数据类型
inspect(transdata0[1:5])

4.计算高频词的单味中药，此时，计算的频次就出来了

# 高频词中药计算
TCM_Freq <- itemFrequency(x = transdata0,"absolute")%>%
  sort(decreasing = TRUE)%>%
  as.data.frame()
TCM_Freq <- TCM_Freq%>%
  mutate("中药名称" = rownames(TCM_Freq), .before = 1)
TCM_Freq 

5.我们开始对这个数据进行绘图，同时使用程序保存图片

# 绘制柱状图
library(ggsci)
TCM_Freq$中药名称<- factor(TCM_Freq$中药名称,levels = TCM_Freq$中药名称)
p1 <- ggplot(TCM_Freq[1:10,],aes(x = reorder(中药名称,.),.,
                            fill = 中药名称,
                            label = .))+
  geom_bar(stat="identity",width = 0.8)+
  coord_flip()+
  scale_fill_aaas()+
  labs(x = "中药名称",y = "频次")+
  scale_y_continuous(limits = c(0,35),expand = c(0,0))+
  geom_label(nudge_y = 2)+
  theme_bw()+
  theme(legend.position = "none",
        axis.title = element_text(size = 12,face = "bold"),
        axis.text = element_text(size = 10,face = "bold"),
        text = element_text(family = "serif"))

tiff("中药词频柱状图.tif",
    width = 11,height = 9,
    units = "cm",res = 300,
    compression = "lzw",)
p1
dev.off()

6.接下来我们使用关联规则进行两味及其以上的组合分析，首先利用算法分析关联规则

# 分析关联规则
myrules <- apriori(transdata0,
                   parameter = list(supp = 0.07,
                                    conf = 0.6,
                                    target = "rules"))

summary(myrules)

7.将结果按照出现的频次进行排序，结果转化为dataframe

# 按照出现的频次进行排序
myrules.conf <- sort(myrules,by="count",decreasing = T)
TCM_FJ <- inspect(myrules.conf)%>%as.data.frame()

8.将dataframe中第一和第二列数据中的括号取消，使用paste0函数将两列粘贴在一起生成新的列

# 去除掉左右的{}
TCM_FJ$lhs <- gsub("[{}]","",TCM_FJ$lhs)
TCM_FJ$rhs <- gsub("[{}]","",TCM_FJ$rhs)
TCM_FJ组合` <- paste0(TCM_FJ$lhs,", ",TCM_FJ$rhs)
TCM_FJ

9.我们会发现，生成的数据中，有些数据是重复的，如第8和9行，为石菖蒲、远志、白芍和白芍、远志、石菖蒲，所以我们保留一个组合就可以了，使用下面函数进行排序并去除掉一个，写成excel文件

# 将组合的列中的中药排序，去除重复内容
TCM_FJ$组合 <- sapply(TCM_FJ$组合,
                    function(x) sort(unlist(str_split(x,", "))))
TCM_FJ <- TCM_FJ[!duplicated(TCM_FJ$组合),]

write.xlsx(TCM_FJ,"temp.xlsx")

10.最后，我们要人工将count列和组合列进行筛选，如两味药的前10名和三味药的前10名，整理出来，如下所示

11.最终使用R语言对其进行绘图即可，保存图片用于发表文章

# 绘图2味药物
TCM_2 <- read.xlsx("temp.xlsx",sheet = 2)
TCM_2$组合<- factor(TCM_2$组合,levels = TCM_2$组合)
p_2 <- ggplot(TCM_2,aes(x = reorder(组合,count),count,
                           fill = 组合,
                           label = count))+
  geom_bar(stat="identity",width = 0.8)+
  coord_flip()+
  scale_fill_npg()+
  labs(x = "配对中药(2味)",y = "频次")+
  scale_y_continuous(limits = c(0,25),expand = c(0,0))+
  geom_label(nudge_y = 2)+
  theme_bw()+
  theme(legend.position = "none",
        axis.title = element_text(size = 12,face = "bold"),
        axis.text = element_text(size = 10,face = "bold"),
        text = element_text(family = "serif"))
tiff("配对中药(2味).tif",
     width = 11,height = 9,
     units = "cm",res = 300,
     compression = "lzw",)
p_2
dev.off()

12.绘制3味药物的图形并进行保存

# 绘图3味药物
TCM_3 <- read.xlsx("temp.xlsx",sheet = 3)
TCM_3$组合<- factor(TCM_3$组合,levels = TCM_3$组合)
p_3 <- ggplot(TCM_3,aes(x = reorder(组合,count),count,
                 fill = 组合,
                 label = count))+
  geom_bar(stat="identity",width = 0.8)+
  coord_flip()+
  scale_fill_npg()+
  labs(x = "配对中药(3味)",y = "频次")+
  scale_y_continuous(limits = c(0,20),expand = c(0,0))+
  geom_label(nudge_y = 1)+
  theme_bw()+
  theme(legend.position = "none",
        axis.title = element_text(size = 12,face = "bold"),
        axis.text = element_text(size = 10,face = "bold"),
        text = element_text(family = "serif"))

tiff("配对中药(3味).tif",
     width = 11,height = 9,
     units = "cm",res = 300,
     compression = "lzw",)
p_3
dev.off()

13.好了，这就是今天讲解的，如何从一组方剂中，挖掘出频次最高的单味、两味、三味中药！