如何用R构建带分类计数的层级树状图？已尝试data.tree遇阻

需求说明 #

需要构建层级图展示数据框中不同分类的计数，层级越往下分类越具体；要求能根据数据框和指定层级变量自动统计各分类的数量，且方案可重复应用于不同数据集。

示例数据 #

tree_data = data.frame(Context = c("urban", "rural", "urban", "urban", "rural", "rural"),
                Lighting = c("daylight", "dark", "dark", "daylight", "daylight", "dark"),
                Driver_age = c("Senior", "Adult", "Adult", "Adult", "Adult", "Senior"))

现有基础操作 #

已使用data.tree包构建基础层级结构：

library(data.tree)

tree_data$pathString = paste("crashes",
                             tree_data$Context,
                             tree_data$Lighting,
                             tree_data$Driver_age,
                             sep = "/")

crashes = as.Node(tree_data)
print(crashes)

后续解决方案 #

1. 为节点添加计数属性 #

data.tree的节点自带totalCount属性，代表该节点下包含的叶子节点总数（即对应分组的观测计数）。可以通过Do()方法将该值赋值为自定义的Count属性，方便后续调用：

# 为每个节点添加Count属性
crashes$Do(function(node) {
  node$Count <- node$totalCount
})

# 查看带计数的节点结构
print(crashes, "Count")

输出示例：

levelName          Count
1 crashes                6
2  ¦--urban              3
3  ¦   ¦--daylight       2
4  ¦   ¦   °--Senior     1
5  ¦   ¦   °--Adult      1
6  ¦   °--dark           1
7  ¦       °--Adult      1
8  °--rural              3
9      ¦--dark           2
10     ¦   °--Adult      1
11     ¦   °--Senior     1
12     °--daylight       1
13         °--Adult      1

2. 可视化层级计数树 #

推荐使用networkD3包生成交互式层级树图，既美观又易于适配不同数据集：

library(networkD3)

# 将data.tree节点转换为networkD3所需的列表格式
tree_list <- ToListExplicit(crashes, unname = TRUE)

# 为节点名称添加计数标注
add_count_label <- function(node) {
  node$name <- paste0(node$name, " (", node$Count, ")")
  if (!is.null(node$children)) {
    node$children <- lapply(node$children, add_count_label)
  }
  return(node)
}
tree_list <- add_count_label(tree_list)

# 生成交互式树图
diagonalNetwork(
  List = tree_list,
  nodeSize = JS("function(d) { return d.Count * 10; }"), # 根据计数调整节点大小
  textColour = "#333",
  opacity = 0.9,
  margin = list(top = 10, right = 100, bottom = 10, left = 100)
)

如果需要静态图，可以使用data.tree自带的plot()函数，配合nodeLabel参数显示计数：

plot(crashes, 
     nodeLabel = function(node) paste(node$name, "\n(", node$Count, ")", sep = ""),
     edgeLabel = NULL)

3. 通用函数封装 #

将上述步骤封装为可复用函数，只需传入数据框和层级变量即可生成带计数的层级树：

library(data.tree)
library(networkD3)
library(jsonlite)

build_count_tree <- function(data, hierarchy_vars) {
  # 构建pathString
  data$pathString <- paste("root", do.call(paste, c(data[hierarchy_vars], sep = "/")), sep = "/")
  # 转换为节点对象
  tree <- as.Node(data)
  # 添加计数属性
  tree$Do(function(node) node$Count <- node$totalCount)
  
  # 处理节点名称，添加计数标注
  tree_list <- ToListExplicit(tree, unname = TRUE)
  add_count_label <- function(node) {
    node$name <- paste0(node$name, " (", node$Count, ")")
    if (!is.null(node$children)) {
      node$children <- lapply(node$children, add_count_label)
    }
    return(node)
  }
  tree_list <- add_count_label(tree_list)
  
  # 生成交互式可视化
  p <- diagonalNetwork(
    List = tree_list,
    nodeSize = JS("function(d) { return d.Count * 10; }"),
    textColour = "#333",
    opacity = 0.9,
    margin = list(top = 10, right = 100, bottom = 10, left = 100)
  )
  
  # 返回树对象和可视化结果
  return(list(tree = tree, plot = p))
}

# 使用示例
result <- build_count_tree(tree_data, c("Context", "Lighting", "Driver_age"))
# 查看树结构
print(result$tree, "Count")
# 显示可视化图
result$plot

内容的提问来源于stack exchange，提问作者Kelly