predict.R

library(forecast)
library(RJSONIO)
library(lubridate)
library(zoo)
library(tidyverse)


if (file.exists('local_config.R')){
  library(RPostgreSQL)
  source('local_config.R')
  
  drv <- dbDriver("PostgreSQL")
  
  con <- dbConnect(drv, dbname="bluelytics")
  dbSendQuery(con,"SET datestyle TO iso")
  rs <- dbSendQuery(con, "select median(value_sell) as value, date_trunc('day', date) as date from dolar_blue_dolarblue where source_id <> 'oficial' and date >= '2013-01-01' group by date_trunc('day', date) order by date;")
  base_data <- fetch(rs,n=-1)
  
  dbDisconnect(con)
  dbUnloadDriver(drv)
}else{
  base_data = read_csv('export_blue.csv')
}

agg <- setNames(aggregate(base_data$value, by=list(as.Date(base_data$date)), FUN=mean), c('date', 'x'))

days <- data.frame ( date = as.Date(seq.POSIXt(as.POSIXlt(min(agg$date)), as.POSIXlt(max(agg$date)), by = "1 day")))

final <- na.locf(merge(days, agg, by="date", all.x=TRUE))
final$x <- as.numeric(final$x)

basedate = min(ymd(final$date))

fts <- ts(final$x, start=c(year(basedate), yday(basedate)), frequency=365)

fit_arima <- Arima(fts, order=c(4,2,3))

horizon <- 60

forecasted <- forecast(fit_arima, h=horizon)

lower_80 <- forecasted$lower[,1]
lower_95 <- forecasted$lower[,2]
upper_80 <- forecasted$upper[,1]
upper_95 <- forecasted$upper[,2]
mean <- as.numeric(forecasted$mean)

data_forecasted = data.frame(lower_95, lower_80, mean, upper_80, upper_95)

dates_history <- seq(basedate, by='day', length.out=nrow(days))
dates_forecasted <- seq(as.Date(tail(dates_history, n=1))+1, by='day', length.out=horizon)

export_history <- strftime(dates_history, '%d/%m/%Y')
export_forecasted <- strftime(dates_forecasted, '%d/%m/%Y')

json_forecasted <- toJSON(data_forecasted)
json_history <- toJSON(data.frame(history=base_data$value))
json_dates_history <- toJSON(export_history)
json_dates_forecasted <- toJSON(export_forecasted)

write(json_forecasted, file.path(path_output, 'json_forecasted.json'))
write(json_history, file.path(path_output, 'json_history.json'))
write(json_dates_history, file.path(path_output, 'json_dates_history.json'))
write(json_dates_forecasted , file.path(path_output, 'json_dates_forecasted.json'))