Project 1

library(readxl)
library(dplyr)
library(forecast)
library(fma)
library(tsoutliers)
library(tidyverse)

data <- read_excel("data.xlsx", sheet="S04", skip=2)
df.1 <- data %>%
          select(c('SeriesInd', 'Var01')) %>%
          filter(SeriesInd<= 43021)  %>%
          rename(ts = Var01)

df.2 <- data %>%
          select(c('SeriesInd', 'Var02')) %>%
          filter(SeriesInd<= 43021)  %>%
          rename(ts = Var02)

data <- read_excel("data.xlsx", sheet="S05", skip=2)
df.3 <- data %>%
          select(c('SeriesInd', 'Var02')) %>%
          filter(SeriesInd<= 43021)  %>%
          rename(ts = Var02) 

df.4 <- data %>%
          select(c('SeriesInd', 'Var03')) %>%
          filter(SeriesInd<= 43021)  %>%
          rename(ts = Var03)

data <- read_excel("data.xlsx", sheet="S06", skip=2)
df.5 <- data %>%
          select(c('SeriesInd', 'Var05')) %>%
          filter(SeriesInd<= 43021)  %>%
          rename(ts = Var05)

df.6 <- data %>%
          select(c('SeriesInd', 'Var07')) %>%
          filter(SeriesInd<= 43021) %>%
          rename(ts = Var07)
all_df <- list(df.1, df.2, df.3, df.4, df.5, df.6)

build_model <- function(train) {
  model <- train %>%
              tsclean() %>%
              auto.arima()
  return(model)
}

create_plot <- function(model, test, var){
  predict1 <- forecast(model, h=488)
  mape <- accuracy(predict1, test)["Test set", "MAPE"]
  title <- sprintf("MAPE of %s is %f" , var, mape)
  predict2 <- forecast(model, h=140)
  return(autoplot(predict2, main=title))
}

iterate <- function(df_list){
  variables <- list("S04 Var01", "S04 Var02", "S05 Var02", "S05 Var03", "S06 Var05", "S06 Var07")
  plot_list = list()
  i <- 1
  for (x in df_list){
    train <- window(x$ts, end=1135)
    test <- window(x$ts, start=1136)
    model <- build_model(train)
    plot_list[[i]] <- create_plot(model, test, variables[i])
    i <- i + 1
  }
return(gridExtra::grid.arrange(grobs = plot_list))
}

iterate(all_df)