我一直在研究 for 回圈與函式系列相比的優缺點，apply()答案并不明確（apply()總是比 for 回圈快可能不正確，具體取決于具體情況）。所以我想根據我的實際資料測驗各種選項。
下面是一個 for 回圈，對我來說看起來很簡單，但我不確定如何用lapply(). 我認為lapply()是正確的，因為 for 回圈會生成一個串列物件。
我需要運行此分析的實際資料是一個包含 250 萬行、30 多列的資料框，因此我想針對各種選項運行速度測驗。
任何解釋都是最有幫助的。我在網上找到的示例解釋簡單，或者 for-loops 示例過于復雜，我希望學習使用apply()族函式，因為它們看起來比 for-loops 非常有用且易于閱讀。
這是簡化的 for 回圈代碼，帶有示例資料框，它可以正確運行以用于示例目的：
# Set up data frame to perform migration analysis on:
data <-
data.frame(
ID = c(1,1,1,2,2,2,3,3,3),
Period = c(1, 2, 3, 1, 2, 3, 1, 2, 3),
Values = c(5, 10, 15, 0, 2, 4, 3, 6, 9),
Flags = c("X0","X1","X2","X0","X2","X0", "X2","X1","X0")
)
# Function to set-up base table:
setTable <- function(data){
df <- data.frame(matrix(NA, ncol=length(unique(data$Flags)), nrow=length(unique(data$Flags))))
row.names(df) <- unique(data$Flags)
names(df) <- unique(data$Flags)
return(df)
}
# Function to complete migration table with for-loop:
migration <- function(data, from=1, to=3){
df <- setTable(data)
for (i in unique(data$ID)){
id_from <- as.character(data$Flags[(data$ID == i & data$Period == from)])
id_to <- as.character(data$Flags[data$ID == i & data$Period == to])
column <- which(names(df) == id_from)
row <- which(row.names(df) == id_to)
df[row, column] <- ifelse(is.na(df[row, column]), 1, df[row, column] 1)
}
return(df)
}
# Now to run the function:
test1 <- migration(data, from=1, to=3)
uj5u.com熱心網友回復：
當談到 R 中的速度時，您幾乎總是可以依靠 library( data.table )：
library(data.table)
DT <- setDT(data.frame(
ID = c(1,1,1,2,2,2,3,3,3,4,4,4),
Period = c(1, 2, 3, 1, 2, 3, 1, 2, 3, 1, 2, 3),
Values = c(5, 10, 15, 0, 2, 4, 3, 6, 9, 3, 6, 9),
Flags = c("X0","X1","X2","X0","X2","X0", "X2","X1","X0", "X2","X1","X0")
))
unique_flags <- unique(DT$Flags)
all_flags <- setDT(expand.grid(list(first_flag = unique_flags, last_flag = unique_flags)))
resultDT <- dcast(DT[, .(first_flag = first(Flags), last_flag = last(Flags)), by = ID][
,.N, c("first_flag", "last_flag")][
all_flags, on = c("first_flag", "last_flag")], last_flag ~ first_flag, value.var = "N")
print(resultDT)
一步步：
library(data.table)
DT <- setDT(data.frame(
ID = c(1,1,1,2,2,2,3,3,3,4,4,4),
Period = c(1, 2, 3, 1, 2, 3, 1, 2, 3, 1, 2, 3),
Values = c(5, 10, 15, 0, 2, 4, 3, 6, 9, 3, 6, 9),
Flags = c("X0","X1","X2","X0","X2","X0", "X2","X1","X0", "X2","X1","X0")
))
unique_flags <- unique(DT$Flags)
all_flags <- setDT(expand.grid(list(first_flag = unique_flags, last_flag = unique_flags)))
resultDT <- DT[, .(first_flag = first(Flags), last_flag = last(Flags)), by = ID] # find relevant flags
resultDT <- resultDT[,.N, c("first_flag", "last_flag")] # count transitions
resultDT <- resultDT[all_flags, on = c("first_flag", "last_flag")] # merge all combinations
resultDT <- dcast(resultDT, last_flag ~ first_flag, value.var = "N") # dcast
print(resultDT)
關于lapply你可以做的（我更喜歡data.table）：
# Set up data frame to perform migration analysis on:
input_data <-
data.frame(
ID = c(1,1,1,2,2,2,3,3,3),
Period = c(1, 2, 3, 1, 2, 3, 1, 2, 3),
Values = c(5, 10, 15, 0, 2, 4, 3, 6, 9),
Flags = c("X0","X1","X2","X0","X2","X0", "X2","X1","X0")
)
# Function to set-up base table:
setTable <- function(data){
DF <- data.frame(matrix(NA, ncol=length(unique(data$Flags)), nrow=length(unique(data$Flags))))
row.names(DF) <- unique(data$Flags)
names(DF) <- unique(data$Flags)
return(DF)
}
# Function to complete migration table with for-loop:
migration <- function(data, from=1, to=3){
DF <- setTable(data)
lapply(seq_along(unique(data$ID)), function(i){
id_from <- as.character(data$Flags[(data$ID == i & data$Period == from)])
id_to <- as.character(data$Flags[data$ID == i & data$Period == to])
column <- which(names(DF) == id_from)
row <- which(row.names(DF) == id_to)
DF[row, column] <<- ifelse(is.na(DF[row, column]), 1, DF[row, column] 1)
})
return(DF)
}
# Now to run the function:
test1 <- migration(input_data, from=1, to=3)
轉載請註明出處，本文鏈接：https://www.uj5u.com/qita/439180.html