如何在R中按分组绘制预测与实际数据对比图表并保存

现有两个数据集：

1. Forecast数据集 #

Forecast=structure(list(sub = c("01.01.01 Automatic switches ARMAT", "01.01.01 Automatic switches ARMAT", 
"01.01.01 Automatic switches ARMAT", "01.01.02 Automatic switches KARAT", 
"01.01.02 Automatic switches KARAT", "01.01.02 Automatic switches KARAT"
), ds = c("01.05.2022", "01.06.2022", "01.07.2022", "01.05.2022", 
"01.06.2022", "01.07.2022"), forecast = c(30696L, 30696L, 30696L, 
1089910L, 1089910L, 1089910L), fct = c(22590L, 29243L, 30521L, 
803535L, 1430316L, 1312260L)), class = "data.frame", row.names = c(NA, 
-6L))

其中sub为分组变量，forecast列是月度预测值，fct（实际值）列是实际数据，预测按sub分组单独执行。

2. manager_forecast数据集 #

manager_forecast=structure(list(sud = c("01.01.01 Automatic switches ARMAT", "01.01.01 Automatic switches ARMAT", 
"01.01.02 Automatic switches KARAT", "01.01.02 Automatic switches KARAT", 
"", ""), ds = c("01.06.2022", "01.07.2022", "01.06.2022", "01.07.2022", 
"", ""), manager_forecast = c(76011.04671, 113534.0746, 1802431.458, 
1399089.059, NA, NA)), class = "data.frame", row.names = c(NA, 
-6L))

该数据集结构类似，manager_forecast列是经理预测的销量。

需求 #

按每个sub分组，绘制展示预测值、实际值、经理预测值相关性的对比图表（类似Excel趋势对比图），并将每个分组的图表保存至C:/1/plots文件夹，求最简实现方法。

步骤1：安装并加载必要工具包 #

install.packages(c("tidyverse", "lubridate"))
library(tidyverse)
library(lubridate)

步骤2：数据预处理与合并 #

修正manager_forecast的列名错误，清理无效空行，再与Forecast按分组和日期合并：

# 清理经理预测数据
manager_clean <- manager_forecast %>%
  rename(sub = sud) %>%
  filter(sub != "", !is.na(ds)) %>%
  mutate(ds = dmy(ds))

# 转换日期格式并合并两个数据集
combined_data <- Forecast %>%
  mutate(ds = dmy(ds)) %>%
  left_join(manager_clean, by = c("sub", "ds"))

步骤3：分组绘图并批量保存 #

按sub拆分数据，循环生成每个分组的趋势对比图并保存到指定路径：

# 自动创建保存目录（不存在则新建）
dir.create("C:/1/plots", recursive = TRUE, showWarnings = FALSE)

# 分组绘图+保存
combined_data %>%
  group_split(sub) %>%
  walk(function(df) {
    # 生成合法文件名（替换特殊字符）
    filename <- str_replace_all(df$sub[1], "[^a-zA-Z0-9]", "_") %>%
      paste0("C:/1/plots/", ., ".png")
    
    # 绘制折线+点的对比图
    p <- ggplot(df, aes(x = ds)) +
      geom_line(aes(y = forecast, color = "模型预测值"), linewidth = 1) +
      geom_line(aes(y = fct, color = "实际值"), linewidth = 1) +
      geom_line(aes(y = manager_forecast, color = "经理预测值"), linewidth = 1, na.rm = TRUE) +
      geom_point(aes(y = forecast), color = "blue") +
      geom_point(aes(y = fct), color = "green") +
      geom_point(aes(y = manager_forecast), color = "red", na.rm = TRUE) +
      labs(title = unique(df$sub), x = "日期", y = "销量", color = "数据类型") +
      theme_minimal() +
      theme(plot.title = element_text(hjust = 0.5))
    
    # 保存图表
    ggsave(filename, p, width = 10, height = 6, dpi = 100)
  })

补充说明 #

• 图表类型：若需要柱状对比，可将geom_line替换为geom_col，并添加position = "dodge"实现并列柱状。
• 空值处理：经理预测仅覆盖6、7月数据，绘图时用na.rm=TRUE自动忽略空值，避免报错。
• 文件名适配：替换分组名称中的特殊字符为下划线，防止保存时出现路径错误。

内容的提问来源于stack exchange，提问作者psysky