Question

我写了一个代码来计算观察到的数据和模拟数据之间的RMSE误差。但是我只想在一月份进行。文本文件的数据在第一列中包含日期，在第二列中包含模拟数据，在第三列中包含观察数据。

数据格式如下：

DATE    cout    rout    coub    cinf
UNITS   m3/s    m3/s    m3/s    m3/s
1981-01-01  292.234 305 0   292.234
1981-01-02  293.152 320 0   293.152
1981-01-03  293.985 324 0   293.985
1981-01-04  295.115 308 0   295.115
1981-01-05  296.579 326 0   296.579
1981-01-06  298.266 344 0   298.266
1981-01-07  300.084 342 0   300.084
1981-01-08  301.945 329 0   301.945
1981-01-09  303.747 357 0   303.747
1981-01-10  305.437 351 0   305.437
1981-01-11  306.967 352 0   306.967
1981-01-12  308.281 382 0   308.28

下面的代码用于计算整个数据集的RMSE，与日期无关：

# Function that returns Root Mean Squared Error

# set the working directory
setwd("D:\\Results\\")

# Get the header 1st line of the data
header <-scan("4001968.txt", nlines=1, what =character())

#Define number of lines to skip, which is 2
y <- read.table("4001968.txt",skip=2,header=F,sep="\t")

# Add the character vector header on as the names component
names(y) <- header

#Function for calculating RMSE
rmse <- function(error)
{
  sqrt(mean(error^2))
}

# Convert characater to numeric
y$cout <- as.numeric(as.character(y$cout)) 
y$rout <- as.numeric(as.character(y$rout)) 
actual <- y$cout
predicted <- y$rout

# Calculate error
error <- actual - predicted

# Invocation of functions
rmse(error)

输出将仅是一月份的单个值。

Answer 1

我发现data.table和lubridate软件包对于处理此类问题非常有用：

# libraries
library(data.table)
library(lubridate)

# Function that returns Root Mean Squared Error

# set the working directory
setwd("D:\\Results\\")

# Get the header 1st line of the data
header <-scan("4001968.txt", nlines=1, what =character())

#Define number of lines to skip, which is 2
y <- read.table("4001968.txt",skip=2,header=F,sep="\t")

# Add the character vector header on as the names component
names(y) <- header

#Function for calculating RMSE
rmse <- function(error)
{
  sqrt(mean(error^2))
}

# Convert characater to numeric
y$cout <- as.numeric(as.character(y$cout)) 
y$rout <- as.numeric(as.character(y$rout)) 
y <- as.data.table(y)

# Calculate error
error <- y[month(DATE)==1, cout-rout]

# Invocation of functions
rmse(error)

每月计算观察和模拟数据之间的RMSE误差

1 个答案: