0

我使用了一些代码，这些代码使用 fable 自动将外部回归器列表传递给 ARIMA()（感谢@seapen 和@Mitchell O'Hara-Wild）。

我想通过差分和转换为对数刻度来转换所有变量。据我了解，如果这是在 model() 中完成的，那么 fable 会自动对预测进行反向转换。

请有人帮我修改下面的代码，以便每个自动生成的变量都以 dlog 形式通过 model() 调用，从而允许进行自动反向转换？

library(tidyverse)
library(fable)
library(BBmisc)
library(progressr)
        
# Load data
    df <- tsibbledata::hh_budget
    
    # Build lists of external regressor combinations
    subsets_list <- function(set, subset_size) {
      combn(set, subset_size) %>%
        BBmisc::convertColsToList() %>%
        unname()
    }
    
    xregs <-
      map(.x = 1:(length(df) - 3), .f = subsets_list, 
          set = colnames(df[4:length(df)])) %>% 
      unlist(recursive = F)
    
    # Construct formulas
    rhs <- map_chr(seq_along(xregs), ~ paste(xregs[[.]], collapse = " + "))
    lhs <- "Debt"
    formulas <- map(paste(lhs, rhs, sep = " ~ "), as.formula)
    
    # Create model specifications
    model_specs <- set_names(map(formulas, ARIMA), formulas)
    
    # Train - test split for model fit
    train <- df %>% filter(Year <= 2014)
    test <- df %>% filter(Year > 2014) 
    
    # Estimate models based on training data (added a progress bar since it takes ~30secs)
    fit <- with_progress(train %>% 
      model(!!!model_specs))
    
    # forecast all models using test data
    fc <- fit %>% 
      forecast(test)

非常感谢任何帮助。

4

0 回答 0