#load libraries
devtools::install_github("ahaeusser/echos")
devtools::install_github("ahaeusser/tscv")
install.packages("fable")
remotes::install_github("tidyverts/fable")
remotes::install_github("tidyverts/tsibble")

#load libraries
library(echos)
library(tscv)
library(dplyr)
library(tsibble)
library(fabletools)
library(fable)
Sys.setlocale("LC_TIME", "C")
#> [1] "C"

library(lubridate)
library(dplyr)
library(tibble)
library(tsibble)

date_decision_made = seq(as.Date("1800/1/1"), as.Date("2020/1/1"),by="day")

date_decision_made <- format(as.Date(date_decision_made), "%Y/%m/%d")

property_damages_in_dollars <- rnorm( 80354,100,10)

final_data <- data.frame(date_decision_made, property_damages_in_dollars)

mut = final_data %>%
    mutate(date_decision_made = as.Date(date_decision_made)) %>%
    group_by(week = format(date_decision_made, "%Y-%m")) %>%
    summarise( total = sum(property_damages_in_dollars, na.rm = TRUE))

#pretend the data has been sampled for "training" and "testing"

sample1 <- c("train","test" )

mut$status<- sample(sample1 , 2641 , replace=TRUE, prob=c(0.7, 0.3))
mut$status <- as.factor(mut$status)
mut$Value = mut$total
mut$total = NULL


ym <- function (..., quiet = FALSE, tz = NULL, locale = Sys.getlocale("LC_TIME")) {
    lubridate:::.parse_xxx(..., orders = "ym", quiet = quiet, tz = tz, locale = locale,
                           truncated = 0)
}

mut$week <- ym(mut$week)

tibble <- as.tibble(mut)


z = tibble %>%
    as_tsibble(index = week)

data = z

# Setup for time series cross validation
n_init <- 2400   # size for training window
n_ahead <- 24    # size for testing window (forecast horizon)
mode <- "slide"  # fixed window approach
n_skip <- 23     # skip 23 observations
n_lag <- 0       # no lag

data <- data %>%
  tscv::split_data(
    n_init = n_init,
    n_ahead = n_ahead,
    mode = mode,
    n_skip = n_skip,
    n_lag = n_lag)

models <- data %>%
  filter(sample == "train") %>%
  model(
    "ESN" = ESN(
      Value,
      inf_crit = "BIC",
      max_lag = 6,
      n_fourier = c(3, 3),
      n_initial = 50,
      n_res = 200,
      scale_inputs = c(-1, 1)),
    "sNaive" = SNAIVE(Value ~ lag("week")))

Warning messages:
1: 11 errors (1 unique) encountered for ESN
[11] cannot coerce type 'closure' to vector of type 'double'
 
2: 11 errors (1 unique) encountered for sNaive
[11] .data contains implicit gaps in time. You should check your data and convert implicit gaps into explicit missing values using `tsibble::fill_gaps()` if required.

data <- data %>% 
     fill_gaps(.full = TRUE)

[models <- data %>%
  filter(sample == "train") %>%
  model(
    "ESN" = ESN(
      Value,
      inf_crit = "BIC",
      max_lag = 6,
      n_fourier = c(3, 3),
      n_initial = 50,
      n_res = 200,
      scale_inputs = c(-1, 1)),
    "sNaive" = SNAIVE(Value ~ lag("week")))

## Detailed report of ESN
models %>%
  select(ESN) %>%
  report()

# Forecast models
fcsts <- models %>%
  forecast(h = n_ahead)

actuals <- data %>%
  filter_index("2019-04-10" ~ .)

fcsts %>%
  autoplot(
    actuals,
    level = NULL,
    size = 1)](url)