R语言：Random Forest前置时间序列数据集预处理求助

针对年尺度半小时时间序列数据集（约200个变量、17000条观测），以下是解决三个预处理问题的可执行R代码方案：

核心需求与对应方案 #

• 用对应日期的每日均值填充NA值
• 批量删除字符列，保留日期型列
• 高效批量计算各列每日均值，避免手动处理

准备工作：加载依赖包 #

# 首次运行需安装包
install.packages(c("dplyr", "lubridate"))

# 加载数据处理与日期工具包
library(dplyr)
library(lubridate)

步骤1：统一日期格式并批量移除字符列 #

先将日期列转换为标准POSIXct格式，再筛选保留日期列与数值列，自动剔除所有字符型列：

处理示例数据 #

# 示例数据集
TIMESTAMP <- c("2019-04-27 17:30:00", "2019-04-27 18:00:00", "2019-04-27 18:30:00", "2019-04-27 19:00:00", "2019-04-27 19:30:00", "2019-04-28 10:00:00", "2019-04-28 10:30:00", "2019-04-28 11:00:00", "2019-04-28 11:30:00", "2019-04-28 12:00:00")
dates<-c("2019-04-27", "2019-04-27", "2019-04-27", "2019-04-27", "2019-04-27", "2019-04-28", "2019-04-28", "2019-04-28", "2019-04-28", "2019-04-28")
ch4_flux <- c(NA, 66.39, 65.39, 64.41, 63.52, 62.76, 62.16,NA, 61.54,61.53)
distance <- c(1000,1000,NA,125.35,1000,NA,1000,5.50,NA,1000)
Tau <-c(0.0322000, 0.0495000, 0.1737616, 0.1772567, NA, 0.1246816, 0.1435230, 0.1098670, NA, NA)
filename<- c("2019-04-27T173000_AIU-2079.ghg","2019-04-27T180000_AIU-2079.ghg", "2019-04-27T183000_AIU-2079.ghg","2019-04-27T190000_AIU-2079.ghg", "2019-04-27T193000_AIU-2079.ghg",NA,  "2019-04-28T100000_AIU-2079.ghg","2019-04-28T103000_AIU-2079.ghg", "2019-04-28T110000_AIU-2079.ghg",NA)

dd<- data.frame(TIMESTAMP, dates, ch4_flux, distance,Tau, filename)

# 转换日期格式+筛选保留日期/数值列
dd_clean <- dd %>%
  mutate(
    TIMESTAMP = ymd_hms(TIMESTAMP),
    dates = ymd(dates)
  ) %>%
  select(where(~inherits(.x, c("POSIXct", "numeric"))))

处理原始数据集（AMPdates2） #

直接对原始数据集执行以下代码：

AMPdates2_clean <- AMPdates2 %>%
  mutate(
    TIMESTAMP = ymd_hms(TIMESTAMP),
    date = ymd(date)
  ) %>%
  select(where(~inherits(.x, c("POSIXct", "numeric"))))

步骤2：批量计算各列每日均值 #

按日期分组，批量计算所有数值列的每日均值：

# 示例数据计算每日均值
daily_means <- dd_clean %>%
  group_by(dates) %>%
  summarise(across(where(is.numeric), ~mean(.x, na.rm = TRUE)), .groups = "drop")

# 原始数据集计算每日均值
daily_means_AMP <- AMPdates2_clean %>%
  group_by(date) %>%
  summarise(across(where(is.numeric), ~mean(.x, na.rm = TRUE)), .groups = "drop")

步骤3：用每日均值填充NA值 #

将原始数据与每日均值表关联，批量替换NA值：

# 填充示例数据的NA
dd_filled <- dd_clean %>%
  left_join(daily_means, by = "dates", suffix = c("", "_mean")) %>%
  mutate(across(where(is.numeric), ~ifelse(is.na(.x), get(paste0(cur_column(), "_mean")), .x))) %>%
  select(-ends_with("_mean"))

# 填充原始数据集的NA
AMPdates2_filled <- AMPdates2_clean %>%
  left_join(daily_means_AMP, by = "date", suffix = c("", "_mean")) %>%
  mutate(across(where(is.numeric), ~ifelse(is.na(.x), get(paste0(cur_column(), "_mean")), .x))) %>%
  select(-ends_with("_mean"))

完整整合代码（示例数据） #

# 加载包
library(dplyr)
library(lubridate)

# 示例数据集
TIMESTAMP <- c("2019-04-27 17:30:00", "2019-04-27 18:00:00", "2019-04-27 18:30:00", "2019-04-27 19:00:00", "2019-04-27 19:30:00", "2019-04-28 10:00:00", "2019-04-28 10:30:00", "2019-04-28 11:00:00", "2019-04-28 11:30:00", "2019-04-28 12:00:00")
dates<-c("2019-04-27", "2019-04-27", "2019-04-27", "2019-04-27", "2019-04-27", "2019-04-28", "2019-04-28", "2019-04-28", "2019-04-28", "2019-04-28")
ch4_flux <- c(NA, 66.39, 65.39, 64.41, 63.52, 62.76, 62.16,NA, 61.54,61.53)
distance <- c(1000,1000,NA,125.35,1000,NA,1000,5.50,NA,1000)
Tau <-c(0.0322000, 0.0495000, 0.1737616, 0.1772567, NA, 0.1246816, 0.1435230, 0.1098670, NA, NA)
filename<- c("2019-04-27T173000_AIU-2079.ghg","2019-04-27T180000_AIU-2079.ghg", "2019-04-27T183000_AIU-2079.ghg","2019-04-27T190000_AIU-2079.ghg", "2019-04-27T193000_AIU-2079.ghg",NA,  "2019-04-28T100000_AIU-2079.ghg","2019-04-28T103000_AIU-2079.ghg", "2019-04-28T110000_AIU-2079.ghg",NA)

dd<- data.frame(TIMESTAMP, dates, ch4_flux, distance,Tau, filename)

# 1. 清洗数据
dd_clean <- dd %>%
  mutate(
    TIMESTAMP = ymd_hms(TIMESTAMP),
    dates = ymd(dates)
  ) %>%
  select(where(~inherits(.x, c("POSIXct", "numeric"))))

# 2. 计算每日均值
daily_means <- dd_clean %>%
  group_by(dates) %>%
  summarise(across(where(is.numeric), ~mean(.x, na.rm = TRUE)), .groups = "drop")

# 3. 填充NA值
dd_filled <- dd_clean %>%
  left_join(daily_means, by = "dates", suffix = c("", "_mean")) %>%
  mutate(across(where(is.numeric), ~ifelse(is.na(.x), get(paste0(cur_column(), "_mean")), .x))) %>%
  select(-ends_with("_mean"))

# 查看处理结果
head(dd_filled)

关键说明 #

• 使用dplyr::across()实现批量操作，高效适配200+变量的大数据集
• lubridate包确保日期格式统一，避免分组计算时出现错误
• 填充逻辑仅针对数值列，日期列全程保留，不参与均值计算

内容的提问来源于stack exchange，提问作者shrimp