r语言 - 组合两个'ts'对象时不相容的数组

Question

我在组合预测函数生成的 2 个时间序列对象时遇到错误。

这是源代码

# For Time Series Analysis
library(timeSeries)
library(forecast)
library(fpp)
library(tseries)
library(TSA)

# For manipulating data
library(magrittr)
library(forcats)
library(dplyr)
library(tidyr)
library(readr)
library(lubridate)
AP <-AirPassengers
splitTrainXvat <- function(tser, perc_train){
ntrain <- floor(length(as.vector(tser)) * perc_train)
nval <- length(as.vector(tser)) - ntrain
ttrain <- ts(as.vector(tser[1:ntrain]), start = start(tser), frequency = frequency(tser))
tval <- ts(as.vector(tser[ntrain + 1:nval]), start = end(ttrain) + deltat(tser), 
frequency = frequency(tser))
stopifnot(length(ttrain) == ntrain)
stopifnot(length(tval) == nval)
list(ttrain, tval)
}
ts_all <- AP
data <- splitTrainXvat(ts_all, 0.95)
data
ts_train <- data[[1]]
ts_val <- data[[2]]
# preparing model with training set 
mod.hw.add <- HoltWinters(ts_train, seasonal = "add")
mod.arima <- auto.arima(ts_train, max.p = 2, max.q = 2, max.d = 2,
max.P = 2, max.Q = 2, max.D = 2, allowdrift = T,
stepwise = F, approximation = F)
mod.arima.boxcox <- auto.arima(ts_train, max.p=2, max.q=2,
max.P=2, max.Q=2, max.d=2, max.D=2, allowdrift = T,
stepwise = F, approximation = F, lambda = 0)
# Preparing model with testing set and forecasting
pred.hw.add <- forecast(mod.hw.add, h = length(ts_val)+8)$mean
pred.arima <- forecast(mod.arima, h = (length(ts_val)+8))$mean 
pred.arima.boxcox <- forecast(mod.arima.boxcox, h = length(ts_val)+4)$mean
# Combining the two forecasts
Y <- (as.matrix(pred.hw.add) + as.matrix(pred.arima.boxcox))

执行此操作时生成的错误是

Y <- (as.matrix(pred.hw.add) + as.matrix(pred.arima.boxcox))

as.matrix(pred.hw.add( + as.matrix(pred.arima.boxcox( 中的错误： 不兼容阵列

知道为什么会这样吗？

Answer 1

您正在尝试将 2 个向量转换为 2 个不同的矩阵，然后添加 - 这是抛出错误，因为两个向量具有不同的长度 - 因此矩阵将具有不同的维度。是否要创建 2 列数据框？给你-

Y <- cbind(pred.hw.add , pred.arima.boxcox)