优化循环速度

时间:2019-04-21 00:35:49

标签: r

以下代码循环遍历vars中的大约120个变量。目前,该文件大约需要80秒才能完成,这很长。我很想知道是否有人在加速我的代码方面看到了低落的果实。我有32个核心,但是为DoParallel更改核心= 32并没有给我带来太多好处。也许将for循环重写为apply?

# rm(list = ls()) #clears the workspace


library(plyr)
library(ggplot2)
library(scales)
library(foreach) 
library(doParallel)
library(forecast)

# set up
options(scipen = 999)  #removes scientific notation
registerDoParallel(cores = 32) 

# read data
proj_path = "P:/R"
Macro <- read.csv("P:/Earnest/Old/R/Input.csv")

# prep data
source("P:/Earnest/Old/R/VarTS.R")  #Calls variables

cbind.fill <- function(...){
  nm <- list(...) 
  nm <- lapply(nm, as.matrix)
  n <- max(sapply(nm, nrow)) 
  do.call(cbind, lapply(nm, function (x) 
    rbind(x, matrix(, n-nrow(x), ncol(x))))) 
}

len <- nrow(Macro)

# Loads all variable names from Macro and Macro2
vars_macro = names(Macro)[!names(Macro) %in% c("qtrs", "y", "s1", "s2", "s3", "date")]  #Returns names in Macro not in "qtrs", "y", "s1", "s2", "s3"
vars_macro2 = names(Macro2)[!names(Macro2) %in% c("y", "s1", "s2", "s3")]
vars_macro3 = names(Macro3)[!names(Macro3) %in% c("y", "s1", "s2", "s3")]
vars_macroall = names(Macroall)[!names(Macroall) %in% c("y", "s1", "s2", "s3")]
vars = c(vars_macro, vars_macro2, vars_macro3)
consensus = Macro[[2]][len]

fit    <- list()
output <- list()
forecast <- list()


for(m in vars){
  thedata <- get(m)
  output[[m]] <- list() # treat output as a list-of-lists 
  fit[[m]] <- list() # treat fit as a list-of-lists

  for(z in rev(1:6)) {
    tryCatch(  #added tryCatch because for loop was breaking on error around number of dimensions
      expr = {
        x <- thedata[1:(len-z),1:1]
        x <- ts((x), start = c(2016, 3), frequency = 4)
        y <- Macro[1:(len-z),2:2]
        y <- ts((y), start = c(2016, 3), frequency = 4)
        t <- thedata[(len+1-z):(len+1-z),1:1]
        t <- ts((t), start = c(2018, 4), frequency = 4) 

        #fit model
        fit[[m]][[z]] <-auto.arima(y,xreg=x,seasonal=TRUE,parallel = TRUE,num.cores = NULL) #D=1 enforces seasonality
        output[[m]][[z]] <- forecast(fit[[m]][[z]],xreg=t)$mean
      },
      error = function(e){
        message("* Caught an error on itertion", m)
        print(e)
      }
    )}}

output
output2 <- matrix(unlist(output), ncol = length(vars))
output2 <- output2[nrow(output2):1,]


matrixoutput <- output2[1:5,1:length(vars)]
matrixactual <- matrix(Macro[(len-5):(len-1),2:2])
matrixactual <- c(matrixactual)


MAPE <- colMeans(abs((matrixactual - matrixoutput) / matrixactual))
Forecast <- output2[6:6,1:length(vars)]
Delta <- ((Forecast - consensus) / consensus)
LastMAPE <- abs(((matrixoutput[5:5,] - matrixactual[5:5])/matrixactual[5:5]))

df <- data.frame(vars,Forecast, MAPE, Delta)
dflast <- data.frame(vars,Forecast, MAPE, LastMAPE, Delta )

df_macro1 = df[df$vars %in% vars_macro,]
df_macro1[17:17,] <- NA 
df <- df[with(df, order(MAPE)), ] #sorts after original variables
df_macro2 = df[df$vars %in% vars_macro2,]
df_macro2$blankVar = NA
df_macro3 = df[df$vars %in% vars_macro3,]
df_macro3$blankVar = NA
df_macroall = df[df$vars %in% vars_macroall,]
df_macroall$blankVar = NA
df_macro4 = df[df$vars %in% names(Macro4),]
df_macro4$blankVar = NA

df_macro5 = dflast[dflast$vars %in% vars_macroall,]
df_macro5 <- df_macro5[with(df_macro5, order(LastMAPE)), ] 
df_macro5$blankVar = NA

df_macro1 <- rbind.fill(df_macro1, df_macro5)

a = cbind.fill(df_macro1, df_macro2, df_macro3, df_macroall, df_macro4, df, consensus)

print(df, row.names = FALSE)
write.csv(a, "P:/Earnest/Old/R/OutputTSList.csv", row.names = F, na="")

不确定是否有帮助,请提供dput(Macro)

dput(Macro)
structure(list(qtrs = structure(1:14, .Label = c("15_Q3", "15_Q4", 
"16_Q1", "16_Q2", "16_Q3", "16_Q4", "17_Q1", "17_Q2", "17_Q3", 
"17_Q4", "18_Q1", "18_Q2", "18_Q3", "QQ_New"), class = "factor"), 
    y = c(121.3, 131.1, 142.5, 156.4, 168.7, 176.2, 177, 186.6, 
    199.6, 208.4, 214, 226.2, 232.5, 233.3), c1372 = c(0.51798059487155, 
    0.605074600778554, 0.70840580233259, 0.796638389230915, 0.800936365512504, 
    0.862507422705653, 0.906587776772603, 0.965869599669482, 
    1.02057681299029, 1.0173665712577, 1.00815541019123, 1.11026857023261, 
    1.12549360199319, 1.17070543044674), c5244 = c(0.0158288398871533, 
    0.0186529717846534, 0.0335038479568057, 0.0322124481706554, 
    0.0473432307176583, 0.0372644166954006, 0.055124227441671, 
    0.0462947124597511, 0.0595395997947759, 0.079813226336006, 
    0.0632338684298483, 0.0359582444979424, 0.0399978873936274, 
    0.0363530147033467), c5640 = c(0.0695411232121069, 0.0711030107194139, 
    0.0768960904393596, 0.0937721113616879, 0.0912072768529112, 
    0.0948627915873504, 0.0898598251896699, 0.102519015439631, 
    0.117307571608132, 0.116512410019832, 0.112621649435311, 
    0.113373707050245, 0.11920732067264, 0.11385677519257), c6164 = c(0.165253620685311, 
    0.180939722142955, 0.204839371353829, 0.230388360169478, 
    0.245455819824873, 0.250222069413121, 0.267517323013963, 
    0.301455130772129, 0.312527568603722, 0.318684362849784, 
    0.336297671149745, 0.385321973576628, 0.393392171202544, 
    0.414026295628249), b1372 = c(0.220276379575007, 0.232259423605283, 
    0.239015099925248, 0.29722406784095, 0.305759227349267, 0.314812674203001, 
    0.373507924872403, 0.376216626537958, 0.450679682818151, 
    0.422160030256414, 0.398670890305128, 0.380896038096525, 
    0.339513818723946, 0.365265284571949), b5244 = c(0.0256963971001724, 
    0.0308736893223339, 0.0314727889765328, 0.0342560993718647, 
    0.0329261690808683, 0.0341169107838618, 0.0500316002161605, 
    0.057066652393088, 0.0637597978553195, 0.102100656473269, 
    0.109515398926193, 0.0509080775409312, 0.034576665601428, 
    0.037353167421955), b5640 = c(0.0610914743954476, 0.0681070468175109, 
    0.0680584203087885, 0.0737178858316377, 0.0657525044040775, 
    0.0634389081514569, 0.0655890933419926, 0.0689747904574716, 
    0.0653176858840394, 0.0683221933318993, 0.0692822163266979, 
    0.0648739229545749, 0.0613089747918081, 0.0681802570906864
    ), b6164 = c(0.106769764392002, 0.117293493937739, 0.128632140410947, 
    0.146139699267301, 0.15999997720466, 0.170137316488036, 0.188733545209946, 
    0.192072924866328, 0.200314760101575, 0.206572493122192, 
    0.21531555211795, 0.187877279779437, 0.161952291803993, 0.160944253061549
    ), v1372 = c(0.00268999015293817, 0.00312395322452212, 0.00339511453015627, 
    0.00345686458302532, 0.00342490795325169, 0.0036875222492476, 
    0.00361618758896395, 0.00355297766248592, 0.00386182842589497, 
    0.00356454140879668, 0.00347191673410727, 0.00363595803623375, 
    0.00374222181870868, 0.00371078757415556), v5244 = c(0.000480051602474059, 
    0.00042588395300854, 0.0005605459198973, 0.000529571165782351, 
    0.00057240403833901, 0.000468179333138233, 0.000653241119295764, 
    0.000455570432040571, 0.000535395675184177, 0.00138501873189088, 
    0.00114060122318986, 0.000320532455933637, 0.000333175133801828, 
    0.000314970929257286), v5640 = c(0.000839152227642805, 0.000878653169087127, 
    0.00086250329626335, 0.000928749230480325, 0.000952705037621405, 
    0.0009145284719627, 0.000862662166602764, 0.000861675351344781, 
    0.000900555099811469, 0.000846321708899047, 0.000869990227889332, 
    0.000889602926022706, 0.000847332392691765, 0.000745645512078392
    ), v6164 = c(0.00174116886286925, 0.00190736857470478, 0.00204206516831707, 
    0.00208434471661794, 0.0022434016778137, 0.00215767033473045, 
    0.00216675412062837, 0.00211218329813293, 0.00222445645154173, 
    0.00210022505915819, 0.00196775099859493, 0.00205851514065652, 
    0.00187858647947073, 0.00198038712502613), bv1372 = c(0.00142360941776777, 
    0.00151188642851632, 0.00157840305106086, 0.00168297006976765, 
    0.00157809381382463, 0.00180299614944991, 0.00285754565464732, 
    0.0026777621396315, 0.00314015380649578, 0.00293231409618566, 
    0.00290686161843522, 0.00248890293023165, 0.00168251123284542, 
    0.00179265933772828), bv5244 = c(0.000582722401914161, 0.000711777965499918, 
    0.000761910243648493, 0.000805854835145839, 0.000736857013245957, 
    0.000833892120648163, 0.0013602408759186, 0.0015050207801102, 
    0.0016936381650882, 0.00295309680744017, 0.0031773319850428, 
    0.00106962198904438, 0.000593441969063344, 0.000574791244792
    ), bv5640 = c(0.00115351432401665, 0.00132428243672085, 0.00136224787475921, 
    0.00141606633583978, 0.00116049625522213, 0.000858609082150378, 
    0.000908098663997447, 0.000935982449156028, 0.000899912473850066, 
    0.000835053614508394, 0.000872837946479594, 0.000833516238462063, 
    0.000726891442062557, 0.000774037355521608), bv6164 = c(0.000926271545004555, 
    0.00105864530300842, 0.00109611375444535, 0.00117219207771791, 
    0.00122508269987305, 0.00135585463133827, 0.00195926581029822, 
    0.00187620455518874, 0.00206572868014085, 0.00213169451258196, 
    0.00205259028597028, 0.00136013066654879, 0.000919533667669498, 
    0.0010609844820593), s1 = c(1L, 0L, 0L, 0L, 1L, 0L, 0L, 0L, 
    1L, 0L, 0L, 0L, 1L, 0L), s2 = c(0L, 1L, 0L, 0L, 0L, 1L, 0L, 
    0L, 0L, 1L, 0L, 0L, 0L, 1L), s3 = c(0L, 0L, 1L, 0L, 0L, 0L, 
    1L, 0L, 0L, 0L, 1L, 0L, 0L, 0L), date = structure(c(1L, 1L, 
    1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), .Label = "2018-09-30", class = "factor")), class = "data.frame", row.names = c(NA, 
-14L))

最后,下面是所有变量的定义方式,即以下是VarTS.R(source(“ P:/Earnest/Old/R/VarTS.R”)#Calls变量

中包含的内容。
#Let's define some variables

c1372 <- Macro["c1372"]
c5244 <- Macro["c5244"]
c5640 <- Macro["c5640"]
c6164 <- Macro["c6164"]
b1372 <- Macro["b1372"]
b5244 <- Macro["b5244"]
b5640 <- Macro["b5640"]
b6164 <- Macro["b6164"]
v1372 <- Macro["v1372"]
v5244 <- Macro["v5244"]
v5640 <- Macro["v5640"]
v6164 <- Macro["v6164"]
bv1372 <- Macro["bv1372"]
bv5244 <- Macro["bv5244"]
bv5640 <- Macro["bv5640"]
bv6164 <- Macro["bv6164"]


X1372C5244C <- rowMeans(Macro[c("c1372", "c5244")])
X1372C5640C <- rowMeans(Macro[c("c1372", "c5640")])
X1372C6164C <- rowMeans(Macro[c("c1372", "c6164")])
X1372C1372B <- rowMeans(Macro[c("c1372", "b1372")])
X1372C5244B <- rowMeans(Macro[c("c1372", "b5244")])
X1372C5640B <- rowMeans(Macro[c("c1372", "b5640")])
X1372C6164B <- rowMeans(Macro[c("c1372", "b6164")])

X5244C5640C <- rowMeans(Macro[c("c5244", "c5640")])
X5244C6164C <- rowMeans(Macro[c("c5244", "c6164")])
X5244C1372B <- rowMeans(Macro[c("c5244", "b1372")])
X5244C5244B <- rowMeans(Macro[c("c5244", "b5244")])
X5244C5640B <- rowMeans(Macro[c("c5244", "b5640")])
X5244C6164B <- rowMeans(Macro[c("c5244", "b6164")])

X5640C6164C <- rowMeans(Macro[c("c5640", "c6164")])
X5640C1372B <- rowMeans(Macro[c("c5640", "b1372")])
X5640C5244B <- rowMeans(Macro[c("c5640", "b5244")])
X5640C5640B <- rowMeans(Macro[c("c5640", "b5640")])
X5640C6164B <- rowMeans(Macro[c("c5640", "b6164")])

X6164C1372B <- rowMeans(Macro[c("c6164", "b1372")])
X6164C5244B <- rowMeans(Macro[c("c6164", "b5244")])
X6164C5640B <- rowMeans(Macro[c("c6164", "b5640")])
X6164C6164B <- rowMeans(Macro[c("c6164", "b6164")])

X1372B5244B <- rowMeans(Macro[c("b1372", "b5244")])
X1372B5640B <- rowMeans(Macro[c("b1372", "b5640")])
X1372B6164B <- rowMeans(Macro[c("b1372", "b6164")])

X5244B5640B <- rowMeans(Macro[c("b5244", "b5640")])
X5244B6164B <- rowMeans(Macro[c("b5244", "b6164")])

X5640B6164B <- rowMeans(Macro[c("b5640", "b6164")])

X1372C5640C1372B <- rowMeans(Macro[c("c1372", "c5640", "b1372" )])
X1372C5640C5640B <- rowMeans(Macro[c("c1372", "c5640", "b5640" )])
X1372C5640C6164B <- rowMeans(Macro[c("c1372", "c5640", "b6164" )])

X1372C6164C1372B <- rowMeans(Macro[c("c1372", "c6164", "b1372" )])
X1372C6164C5640B <- rowMeans(Macro[c("c1372", "c6164", "b5640" )])
X1372C6164C6164B <- rowMeans(Macro[c("c1372", "c6164", "b6164" )])

X6164C5640C1372B <- rowMeans(Macro[c("c6164", "c5640", "b1372" )])
X6164C5640C5640B <- rowMeans(Macro[c("c6164", "c5640", "b5640" )])
X6164C5640C6164B <- rowMeans(Macro[c("c6164", "c5640", "b6164" )])

XAll3Card <- rowMeans(Macro[c("c1372","c5640","c6164")])
XAll3Card1372B <- rowMeans(Macro[c("c1372","c5640","c6164","b1372")])
XAll3Card5640B <- rowMeans(Macro[c("c1372","c5640","c6164","b5640")])
XAll3Card6164B <- rowMeans(Macro[c("c1372","c5640","c6164","b6164")])

XAll3Bank <- rowMeans(Macro[c("b1372","b5640","b6164")])
XAll3Bank1372C <- rowMeans(Macro[c("b1372","b5640","b6164","c1372")])
XAll3Bank5640C <- rowMeans(Macro[c("b1372","b5640","b6164","c5640")])
XAll3Bank6164C <- rowMeans(Macro[c("b1372","b5640","b6164","c6164")])

XAll13725640 <- rowMeans(Macro[c("c1372", "b1372", "c5640", "b5640")])
XAll13726164 <- rowMeans(Macro[c("c1372", "b1372", "c6164", "b6164")])
XAll56406164 <- rowMeans(Macro[c("c5640", "b5640", "c6164", "b6164")])

XAll4Card <- rowMeans(Macro[c("c1372", "c5244", "c5640", "c6164")])
XAll4Bank <- rowMeans(Macro[c("b1372", "b5244", "b5640", "b6164")])
XAll6 <- rowMeans(Macro[c("c1372","c5640","c6164","b1372","b5640","b6164")])
XAll8 <- rowMeans(Macro[c("c1372", "c5244", "c5640", "c6164","b1372", "b5244", "b5640", "b6164")])





X1372V5244V <- rowMeans(Macro[c("v1372", "v5244")])
X1372V5640V <- rowMeans(Macro[c("v1372", "v5640")])
X1372V6164V <- rowMeans(Macro[c("v1372", "v6164")])
X1372V1372BV <- rowMeans(Macro[c("v1372", "bv1372")])
X1372V5244BV <- rowMeans(Macro[c("v1372", "bv5244")])
X1372V5640BV <- rowMeans(Macro[c("v1372", "bv5640")])
X1372V6164BV <- rowMeans(Macro[c("v1372", "bv6164")])

X5244V5640V <- rowMeans(Macro[c("v5244", "v5640")])
X5244V6164V <- rowMeans(Macro[c("v5244", "v6164")])
X5244V1372BV <- rowMeans(Macro[c("v5244", "bv1372")])
X5244V5244BV <- rowMeans(Macro[c("v5244", "bv5244")])
X5244V5640BV <- rowMeans(Macro[c("v5244", "bv5640")])
X5244V6164BV <- rowMeans(Macro[c("v5244", "bv6164")])

X5640V6164V <- rowMeans(Macro[c("v5640", "v6164")])
X5640V1372BV <- rowMeans(Macro[c("v5640", "bv1372")])
X5640V5244BV <- rowMeans(Macro[c("v5640", "bv5244")])
X5640V5640BV <- rowMeans(Macro[c("v5640", "bv5640")])
X5640V6164BV <- rowMeans(Macro[c("v5640", "bv6164")])

X6164V1372BV <- rowMeans(Macro[c("v6164", "bv1372")])
X6164V5244BV <- rowMeans(Macro[c("v6164", "bv5244")])
X6164V5640BV <- rowMeans(Macro[c("v6164", "bv5640")])
X6164V6164BV <- rowMeans(Macro[c("v6164", "bv6164")])

X1372BV5244BV <- rowMeans(Macro[c("bv1372", "bv5244")])
X1372BV5640BV <- rowMeans(Macro[c("bv1372", "bv5640")])
X1372BV6164BV <- rowMeans(Macro[c("bv1372", "bv6164")])

X5244BV5640BV <- rowMeans(Macro[c("bv5244", "bv5640")])
X5244BV6164BV <- rowMeans(Macro[c("bv5244", "bv6164")])

X5640BV6164BV <- rowMeans(Macro[c("bv5640", "bv6164")])

X1372V5640V1372BV <- rowMeans(Macro[c("v1372", "v5640", "bv1372" )])
X1372V5640V5640BV <- rowMeans(Macro[c("v1372", "v5640", "bv5640" )])
X1372V5640V6164BV <- rowMeans(Macro[c("v1372", "v5640", "bv6164" )])

X1372V6164V1372BV <- rowMeans(Macro[c("v1372", "v6164", "bv1372" )])
X1372V6164V5640BV <- rowMeans(Macro[c("v1372", "v6164", "bv5640" )])
X1372V6164V6164BV <- rowMeans(Macro[c("v1372", "v6164", "bv6164" )])

X6164V5640V1372BV <- rowMeans(Macro[c("v6164", "v5640", "bv1372" )])
X6164V5640V5640BV <- rowMeans(Macro[c("v6164", "v5640", "bv5640" )])
X6164V5640V6164BV <- rowMeans(Macro[c("v6164", "v5640", "bv6164" )])

XAll3CardV <- rowMeans(Macro[c("v1372","v5640","v6164")])
XAll3Card1372BV <- rowMeans(Macro[c("v1372","v5640","v6164","bv1372")])
XAll3Card5640BV <- rowMeans(Macro[c("v1372","v5640","v6164","bv5640")])
XAll3Card6164BV <- rowMeans(Macro[c("v1372","v5640","v6164","bv6164")])

XAll3BankV <- rowMeans(Macro[c("bv1372","bv5640","bv6164")])
XAll3Bank1372V <- rowMeans(Macro[c("bv1372","bv5640","bv6164","v1372")])
XAll3Bank5640V <- rowMeans(Macro[c("bv1372","bv5640","bv6164","v5640")])
XAll3Bank6164V <- rowMeans(Macro[c("bv1372","bv5640","bv6164","v6164")])


XAll13725640V <- rowMeans(Macro[c("v1372", "bv1372", "v5640", "bv5640")])
XAll13726164V <- rowMeans(Macro[c("v1372", "bv1372", "v6164", "bv6164")])
XAll56406164V <- rowMeans(Macro[c("v5640", "bv5640", "v6164", "bv6164")])

XAll4CardV <- rowMeans(Macro[c("v1372", "v5244", "v5640", "v6164")])
XAll4BankV <- rowMeans(Macro[c("bv1372", "bv5244", "bv5640", "bv6164")])
XAll6V <- rowMeans(Macro[c("v1372","v5640","v6164","bv1372","bv5640","bv6164")])
XAll8V <- rowMeans(Macro[c("v1372", "v5244", "v5640", "v6164","bv1372", "bv5244", "bv5640", "bv6164")])




X1372C5244C <- data.frame(X1372C5244C)
X1372C5640C <- data.frame(X1372C5640C)
X1372C6164C <- data.frame(X1372C6164C)
X1372C1372B <- data.frame(X1372C1372B)
X1372C5244B <- data.frame(X1372C5244B)
X1372C5640B <- data.frame(X1372C5640B)
X1372C6164B <- data.frame(X1372C6164B)

X5244C5640C <- data.frame(X5244C5640C)
X5244C6164C <- data.frame(X5244C6164C)
X5244C1372B <- data.frame(X5244C1372B)
X5244C5244B <- data.frame(X5244C5244B)
X5244C5640B <- data.frame(X5244C5640B)
X5244C6164B <- data.frame(X5244C6164B)

X5640C6164C <- data.frame(X5640C6164C)
X5640C1372B <- data.frame(X5640C1372B)
X5640C5244B <- data.frame(X5640C5244B)
X5640C5640B <- data.frame(X5640C5640B)
X5640C6164B <- data.frame(X5640C6164B)

X6164C1372B <- data.frame(X6164C1372B)
X6164C5244B <- data.frame(X6164C5244B)
X6164C5640B <- data.frame(X6164C5640B)
X6164C6164B <- data.frame(X6164C6164B)

X1372B5244B <- data.frame(X1372B5244B)
X1372B5640B <- data.frame(X1372B5640B)
X1372B6164B <- data.frame(X1372B6164B)

X5244B5640B <- data.frame(X5244B5640B)
X5244B6164B <- data.frame(X5244B6164B)

X5640B6164B <- data.frame(X5640B6164B)

X1372C5640C1372B <- data.frame(X1372C5640C1372B)
X1372C5640C5640B <- data.frame(X1372C5640C5640B)
X1372C5640C6164B <- data.frame(X1372C5640C6164B)

X1372C6164C1372B <- data.frame(X1372C6164C1372B)
X1372C6164C5640B <- data.frame(X1372C6164C5640B)
X1372C6164C6164B <- data.frame(X1372C6164C6164B)

X6164C5640C1372B <- data.frame(X6164C5640C1372B)
X6164C5640C5640B <- data.frame(X6164C5640C5640B)
X6164C5640C6164B <- data.frame(X6164C5640C6164B)

XAll3Card <- data.frame(XAll3Card)
XAll3Card1372B <- data.frame(XAll3Card1372B)
XAll3Card5640B <- data.frame(XAll3Card5640B)
XAll3Card6164B <- data.frame(XAll3Card6164B)

XAll3Bank <- data.frame(XAll3Bank)
XAll3Bank1372C <- data.frame(XAll3Bank1372C)
XAll3Bank5640C <- data.frame(XAll3Bank5640C)
XAll3Bank6164C <- data.frame(XAll3Bank6164C)

XAll13725640 <- data.frame(XAll13725640)
XAll13726164 <- data.frame(XAll13726164)
XAll56406164 <- data.frame(XAll56406164)

XAll4Card <- data.frame(XAll4Card)
XAll4Bank <- data.frame(XAll4Bank)
XAll6 <- data.frame(XAll6)
XAll8 <- data.frame(XAll8)




X1372V5244V <- data.frame(X1372V5244V)
X1372V5640V <- data.frame(X1372V5640V)
X1372V6164V <- data.frame(X1372V6164V)
X1372V1372BV <- data.frame(X1372V1372BV)
X1372V5244BV <- data.frame(X1372V5244BV)
X1372V5640BV <- data.frame(X1372V5640BV)
X1372V6164BV <- data.frame(X1372V6164BV)

X5244V5640V <- data.frame(X5244V5640V)
X5244V6164V <- data.frame(X5244V6164V)
X5244V1372BV <- data.frame(X5244V1372BV)
X5244V5244BV <- data.frame(X5244V5244BV)
X5244V5640BV <- data.frame(X5244V5640BV)
X5244V6164BV <- data.frame(X5244V6164BV)

X5640V6164V <- data.frame(X5640V6164V)
X5640V1372BV <- data.frame(X5640V1372BV)
X5640V5244BV <- data.frame(X5640V5244BV)
X5640V5640BV <- data.frame(X5640V5640BV)
X5640V6164BV <- data.frame(X5640V6164BV)

X6164V1372BV <- data.frame(X6164V1372BV)
X6164V5244BV <- data.frame(X6164V5244BV)
X6164V5640BV <- data.frame(X6164V5640BV)
X6164V6164BV <- data.frame(X6164V6164BV)

X1372BV5244BV <- data.frame(X1372BV5244BV)
X1372BV5640BV <- data.frame(X1372BV5640BV)
X1372BV6164BV <- data.frame(X1372BV6164BV)

X5244BV5640BV <- data.frame(X5244BV5640BV)
X5244BV6164BV <- data.frame(X5244BV6164BV)

X5640BV6164BV <- data.frame(X5640BV6164BV)

X1372V5640V1372BV <- data.frame(X1372V5640V1372BV)
X1372V5640V5640BV <- data.frame(X1372V5640V5640BV)
X1372V5640V6164BV <- data.frame(X1372V5640V6164BV)

X1372V6164V1372BV <- data.frame(X1372V6164V1372BV)
X1372V6164V5640BV <- data.frame(X1372V6164V5640BV)
X1372V6164V6164BV <- data.frame(X1372V6164V6164BV)

X6164V5640V1372BV <- data.frame(X6164V5640V1372BV)
X6164V5640V5640BV <- data.frame(X6164V5640V5640BV)
X6164V5640V6164BV <- data.frame(X6164V5640V6164BV)

XAll3CardV <- data.frame(XAll3CardV)
XAll3Card1372BV <- data.frame(XAll3Card1372BV)
XAll3Card5640BV <- data.frame(XAll3Card5640BV)
XAll3Card6164BV <- data.frame(XAll3Card6164BV)

XAll3BankV <- data.frame(XAll3BankV)
XAll3Bank1372V <- data.frame(XAll3Bank1372V)
XAll3Bank5640V <- data.frame(XAll3Bank5640V)
XAll3Bank6164V <- data.frame(XAll3Bank6164V)

XAll13725640V <- data.frame(XAll13725640V)
XAll13726164V <- data.frame(XAll13726164V)
XAll56406164V <- data.frame(XAll56406164V)

XAll4CardV <- data.frame(XAll4CardV)
XAll4BankV <- data.frame(XAll4BankV)
XAll6V <- data.frame(XAll6V)
XAll8V <- data.frame(XAll8V)


s1 <- Macro["s1"]
s2 <- Macro["s2"]
s3 <- Macro["s3"]
y <- Macro["y"]

Macro2 <- data.frame(y, X1372C5244C, X1372C5640C, X1372C6164C, X1372C1372B, X1372C5244B, X1372C5640B, X1372C6164B, 
                     X5244C5640C, X5244C6164C, X5244C1372B, X5244C5244B, X5244C5640B, X5244C6164B, X5640C6164C, 
                     X5640C1372B, X5640C5244B, X5640C5640B, X5640C6164B, X6164C1372B, X6164C5244B, X6164C5640B, X6164C6164B, X1372B5244B, X1372B5640B, X1372B6164B, X5244B5640B, X5244B6164B, X5640B6164B, X1372C5640C1372B, X1372C5640C5640B, X1372C5640C6164B, X1372C6164C1372B, X1372C6164C5640B, X1372C6164C6164B, X6164C5640C1372B, X6164C5640C5640B, X6164C5640C6164B, 
                     XAll3Card, XAll3Card1372B, XAll3Card5640B, XAll3Card6164B, XAll3Bank, XAll3Bank1372C, XAll3Bank5640C, XAll3Bank6164C, XAll13725640, XAll13726164, XAll56406164, XAll4Card, XAll4Bank, XAll6, XAll8, s1, s2, s3)




Macro3 <- data.frame(y, X1372V5244V, X1372V5640V, X1372V6164V, X1372V1372BV, X1372V5244BV, X1372V5640BV, X1372V6164BV, 
                     X5244V5640V, X5244V6164V, X5244V1372BV, X5244V5244BV, X5244V5640BV, X5244V6164BV, X5640V6164V, 
                     X5640V1372BV, X5640V5244BV, X5640V5640BV, X5640V6164BV, X6164V1372BV, X6164V5244BV, X6164V5640BV, X6164V6164BV, X1372BV5244BV, X1372BV5640BV, X1372BV6164BV, X5244BV5640BV, X5244BV6164BV, X5640BV6164BV, X1372V5640V1372BV, X1372V5640V5640BV, X1372V5640V6164BV, X1372V6164V1372BV, X1372V6164V5640BV, X1372V6164V6164BV, X6164V5640V1372BV, X6164V5640V5640BV, X6164V5640V6164BV, 
                     XAll3CardV, XAll3Card1372BV, XAll3Card5640BV, XAll3Card6164BV, XAll3BankV, XAll3Bank1372V, XAll3Bank5640V, XAll3Bank6164V, XAll13725640V, XAll13726164V, XAll56406164V, XAll4CardV, XAll4BankV, XAll6V, XAll8V, s1, s2, s3)

Macro4 <- data.frame(c5640, c6164, b5640, b6164, v5640, v6164, bv5640, bv6164,
                     X5640C6164C, X5640C5640B, X5640C6164B, X6164C5640B, X6164C6164B, X5640B6164B, X6164C5640C5640B, X6164C5640C6164B, XAll56406164, 
                     X5640V6164V, X5640V5640BV, X5640V6164BV, X6164V5640BV, X6164V6164BV, X5640BV6164BV, X6164V5640V5640BV, X6164V5640V6164BV, XAll56406164V)


Macroall <- data.frame(y, c1372, c5640, c6164, b1372, b5640, b6164, v1372, v5640, v6164, bv1372, bv5640, bv6164,
                       X1372C5640C, X1372C6164C, X1372C1372B, X1372C5640B, X1372C6164B, 
                       X5640C6164C, 
                       X5640C1372B, X5640C5640B, X5640C6164B, X6164C1372B, X6164C5640B, X6164C6164B, X1372B5640B, X1372B6164B, X5640B6164B, X1372C5640C1372B, X1372C5640C5640B, X1372C5640C6164B, X1372C6164C1372B, X1372C6164C5640B, X1372C6164C6164B, X6164C5640C1372B, X6164C5640C5640B, X6164C5640C6164B, 
                       XAll3Card, XAll3Card1372B, XAll3Card5640B, XAll3Card6164B, XAll3Bank, XAll3Bank1372C, XAll3Bank5640C, XAll3Bank6164C, XAll13725640, XAll13726164, XAll56406164, XAll6,
                       X1372V5640V, X1372V6164V, X1372V1372BV, X1372V5640BV, X1372V6164BV, 
                       X5640V6164V, 
                       X5640V1372BV, X5640V5640BV, X5640V6164BV, X6164V1372BV, X6164V5640BV, X6164V6164BV, X1372BV5640BV, X1372BV6164BV, X5640BV6164BV, X1372V5640V1372BV, X1372V5640V5640BV, X1372V5640V6164BV, X1372V6164V1372BV, X1372V6164V5640BV, X1372V6164V6164BV, X6164V5640V1372BV, X6164V5640V5640BV, X6164V5640V6164BV, 
                       XAll3CardV, XAll3Card1372BV, XAll3Card5640BV, XAll3Card6164BV, XAll3BankV, XAll3Bank1372V, XAll3Bank5640V, XAll3Bank6164V, XAll13725640V, XAll13726164V, XAll56406164V, XAll6V, s1, s2, s3)

1 个答案:

答案 0 :(得分:1)

auto.arima占用了大部分时间,这意味着没有太多优化。每个auto.arima大约需要200毫秒。 1:6乘以16列乘以200毫秒即为19.2秒。在我的机器上,此解决方案在不启用任何并行内核的情况下要花费大约22秒才能运行。开销不要太寒酸。 最终编辑:使用future_lapply,我在2核i3上将其从22秒提高到了12秒。拥有32核,您应该看起来还不错。

最大的变化是,我为每个时间序列循环(即Macro)对您的rev(1:6)进行了一次子集设置。这有助于避免为每个变量名重复y子集。

然后,使用lapply而不是显式循环,该循环应该比for循环快。

最后一点,我仍然无法运行您的代码,因此不确定这些是否是您想要的结果。

library(future.apply)
plan(multiprocess) 
all_outputs <- lapply(6:1,
                      function (z) {
                        x1 <- ts(Macro[-z, vars], start = c(2016, 3), frequency = 4)
                        y1 <- ts(Macro[-z, 'y'], start = c(2016, 3), frequency = 4)
                        t1 <- ts(Macro[nrow(Macro)-z+1, vars], start = c(2018, 4), frequency = 4)

                        fits <-
                          future_lapply(x1, function(x)
                            auto.arima(
                              y = y1,
                              xreg = x,
                              seasonal = TRUE
                            ))
                        outputs <-
                          lapply(vars,  function(x)
                            x = forecast(fits[[x]], xreg = t1[, x])$mean)

                        names(outputs) <- vars

                        outputs
                      }
                    )

#if you have data.table
data.table::rbindlist(all_outputs)
#or base works fine as well
do.call(rbind, all_outputs)

edit2:结果

> do.call(rbind, all_outputs)
     c1372    c5244    c5640    c6164    b1372    b5244    b5640   
[1,] 207.3153 243.6196 240.3115 195.4832 244.7843 231.6605 236.5413
[2,] 206.3054 241.976  241.5821 198.7039 243.2236 239.3933 239.69  
[3,] 204.1035 244.3271 244.2875 218.5334 243.7765 239.1547 252.0341
[4,] 222.5313 242.494  234.2466 229.7297 242.8088 247.4714 242.9732
[5,] 228.2269 242.7452 222.5183 233.2577 242.0669 242.7832 242.8417
[6,] 237.058  241.8235 235.3313 242.1377 241.7233 241.8443 241.8192
     b6164    v1372    v5244    v5640    v6164    bv1372   bv5244  
[1,] 244.097  243.0404 242.4219 250.7803 245.2109 243.1929 230.8454
[2,] 245.59   239.6461 240.0976 249.8216 244.076  240.2225 239.4011
[3,] 248.9123 240.3455 241.308  256.1201 241.8166 239.2862 238.6731
[4,] 245.1561 240.3136 246.373  257.4908 243.991  241.2637 247.0656
[5,] 242.3743 241.8402 242.5925 252.1454 239.7479 242.9514 242.6109
[6,] 241.8072 241.3552 244.5711 237.4899 241.7374 241.8847 244.9614
     bv5640   bv6164  
[1,] 231.3937 241.8475
[2,] 231.5027 239.3111
[3,] 230.2004 239.2619
[4,] 230.8542 231.8769
[5,] 233.1609 236.7501
[6,] 232.2402 241.8176

修改:数据

library(scales)
library(foreach) 
library(doParallel)
library(forecast)
library(future.apply)

plan(multiprocess) #for future_lapply

Macro <-
  structure(
    list(
      qtrs = structure(
        1:14,
        .Label = c(
          "15_Q3",
          "15_Q4",
          "16_Q1",
          "16_Q2",
          "16_Q3",
          "16_Q4",
          "17_Q1",
          "17_Q2",
          "17_Q3",
          "17_Q4",
          "18_Q1",
          "18_Q2",
          "18_Q3",
          "QQ_New"
        ),
        class = "factor"
      ),
      y = c(
        121.3,
        131.1,
        142.5,
        156.4,
        168.7,
        176.2,
        177,
        186.6,
        199.6,
        208.4,
        214,
        226.2,
        232.5,
        233.3
      ),
      c1372 = c(
        0.51798059487155,
        0.605074600778554,
        0.70840580233259,
        0.796638389230915,
        0.800936365512504,
        0.862507422705653,
        0.906587776772603,
        0.965869599669482,
        1.02057681299029,
        1.0173665712577,
        1.00815541019123,
        1.11026857023261,
        1.12549360199319,
        1.17070543044674
      ),
      c5244 = c(
        0.0158288398871533,
        0.0186529717846534,
        0.0335038479568057,
        0.0322124481706554,
        0.0473432307176583,
        0.0372644166954006,
        0.055124227441671,
        0.0462947124597511,
        0.0595395997947759,
        0.079813226336006,
        0.0632338684298483,
        0.0359582444979424,
        0.0399978873936274,
        0.0363530147033467
      ),
      c5640 = c(
        0.0695411232121069,
        0.0711030107194139,
        0.0768960904393596,
        0.0937721113616879,
        0.0912072768529112,
        0.0948627915873504,
        0.0898598251896699,
        0.102519015439631,
        0.117307571608132,
        0.116512410019832,
        0.112621649435311,
        0.113373707050245,
        0.11920732067264,
        0.11385677519257
      ),
      c6164 = c(
        0.165253620685311,
        0.180939722142955,
        0.204839371353829,
        0.230388360169478,
        0.245455819824873,
        0.250222069413121,
        0.267517323013963,
        0.301455130772129,
        0.312527568603722,
        0.318684362849784,
        0.336297671149745,
        0.385321973576628,
        0.393392171202544,
        0.414026295628249
      ),
      b1372 = c(
        0.220276379575007,
        0.232259423605283,
        0.239015099925248,
        0.29722406784095,
        0.305759227349267,
        0.314812674203001,
        0.373507924872403,
        0.376216626537958,
        0.450679682818151,
        0.422160030256414,
        0.398670890305128,
        0.380896038096525,
        0.339513818723946,
        0.365265284571949
      ),
      b5244 = c(
        0.0256963971001724,
        0.0308736893223339,
        0.0314727889765328,
        0.0342560993718647,
        0.0329261690808683,
        0.0341169107838618,
        0.0500316002161605,
        0.057066652393088,
        0.0637597978553195,
        0.102100656473269,
        0.109515398926193,
        0.0509080775409312,
        0.034576665601428,
        0.037353167421955
      ),
      b5640 = c(
        0.0610914743954476,
        0.0681070468175109,
        0.0680584203087885,
        0.0737178858316377,
        0.0657525044040775,
        0.0634389081514569,
        0.0655890933419926,
        0.0689747904574716,
        0.0653176858840394,
        0.0683221933318993,
        0.0692822163266979,
        0.0648739229545749,
        0.0613089747918081,
        0.0681802570906864
      ),
      b6164 = c(
        0.106769764392002,
        0.117293493937739,
        0.128632140410947,
        0.146139699267301,
        0.15999997720466,
        0.170137316488036,
        0.188733545209946,
        0.192072924866328,
        0.200314760101575,
        0.206572493122192,
        0.21531555211795,
        0.187877279779437,
        0.161952291803993,
        0.160944253061549
      ),
      v1372 = c(
        0.00268999015293817,
        0.00312395322452212,
        0.00339511453015627,
        0.00345686458302532,
        0.00342490795325169,
        0.0036875222492476,
        0.00361618758896395,
        0.00355297766248592,
        0.00386182842589497,
        0.00356454140879668,
        0.00347191673410727,
        0.00363595803623375,
        0.00374222181870868,
        0.00371078757415556
      ),
      v5244 = c(
        0.000480051602474059,
        0.00042588395300854,
        0.0005605459198973,
        0.000529571165782351,
        0.00057240403833901,
        0.000468179333138233,
        0.000653241119295764,
        0.000455570432040571,
        0.000535395675184177,
        0.00138501873189088,
        0.00114060122318986,
        0.000320532455933637,
        0.000333175133801828,
        0.000314970929257286
      ),
      v5640 = c(
        0.000839152227642805,
        0.000878653169087127,
        0.00086250329626335,
        0.000928749230480325,
        0.000952705037621405,
        0.0009145284719627,
        0.000862662166602764,
        0.000861675351344781,
        0.000900555099811469,
        0.000846321708899047,
        0.000869990227889332,
        0.000889602926022706,
        0.000847332392691765,
        0.000745645512078392
      ),
      v6164 = c(
        0.00174116886286925,
        0.00190736857470478,
        0.00204206516831707,
        0.00208434471661794,
        0.0022434016778137,
        0.00215767033473045,
        0.00216675412062837,
        0.00211218329813293,
        0.00222445645154173,
        0.00210022505915819,
        0.00196775099859493,
        0.00205851514065652,
        0.00187858647947073,
        0.00198038712502613
      ),
      bv1372 = c(
        0.00142360941776777,
        0.00151188642851632,
        0.00157840305106086,
        0.00168297006976765,
        0.00157809381382463,
        0.00180299614944991,
        0.00285754565464732,
        0.0026777621396315,
        0.00314015380649578,
        0.00293231409618566,
        0.00290686161843522,
        0.00248890293023165,
        0.00168251123284542,
        0.00179265933772828
      ),
      bv5244 = c(
        0.000582722401914161,
        0.000711777965499918,
        0.000761910243648493,
        0.000805854835145839,
        0.000736857013245957,
        0.000833892120648163,
        0.0013602408759186,
        0.0015050207801102,
        0.0016936381650882,
        0.00295309680744017,
        0.0031773319850428,
        0.00106962198904438,
        0.000593441969063344,
        0.000574791244792
      ),
      bv5640 = c(
        0.00115351432401665,
        0.00132428243672085,
        0.00136224787475921,
        0.00141606633583978,
        0.00116049625522213,
        0.000858609082150378,
        0.000908098663997447,
        0.000935982449156028,
        0.000899912473850066,
        0.000835053614508394,
        0.000872837946479594,
        0.000833516238462063,
        0.000726891442062557,
        0.000774037355521608
      ),
      bv6164 = c(
        0.000926271545004555,
        0.00105864530300842,
        0.00109611375444535,
        0.00117219207771791,
        0.00122508269987305,
        0.00135585463133827,
        0.00195926581029822,
        0.00187620455518874,
        0.00206572868014085,
        0.00213169451258196,
        0.00205259028597028,
        0.00136013066654879,
        0.000919533667669498,
        0.0010609844820593
      ),
      s1 = c(1L, 0L, 0L, 0L, 1L, 0L, 0L, 0L,
             1L, 0L, 0L, 0L, 1L, 0L),
      s2 = c(0L, 1L, 0L, 0L, 0L, 1L, 0L,
             0L, 0L, 1L, 0L, 0L, 0L, 1L),
      s3 = c(0L, 0L, 1L, 0L, 0L, 0L,
             1L, 0L, 0L, 0L, 1L, 0L, 0L, 0L),
      date = structure(
        c(1L, 1L,
          1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L),
        .Label = "2018-09-30",
        class = "factor"
      )
    ),
    class = "data.frame",
    row.names = c(NA,-14L)
  )

# Loads variables
vars_macro <- c("c1372", "c5244", "c5640", "c6164", "b1372", "b5244", "b5640", 
                  "b6164", "v1372", "v5244", "v5640", "v6164", "bv1372", "bv5244", 
                  "bv5640", "bv6164")
vars = c(vars_macro)

all_outputs <- lapply(6:1,
                      function (z) {
                        x1 <- ts(Macro[-z, vars], start = c(2016, 3), frequency = 4)
                        y1 <- ts(Macro[-z, 'y'], start = c(2016, 3), frequency = 4)
                        t1 <- ts(Macro[nrow(Macro)-z+1, vars], start = c(2018, 4), frequency = 4)

                        fits <-
                          future_lapply(x1, function(x)
                            auto.arima(
                              y = y1,
                              xreg = x,
                              seasonal = TRUE
                            ))
                        outputs <-
                          lapply(vars,  function(x)
                            x = forecast(fits[[x]], xreg = t1[, x])$mean)

                        names(outputs) <- vars

                        outputs
                      }
                    )

data.table::rbindlist(all_outputs)
do.call(rbind, all_outputs)