full_forecast_model.Rmd

---
params:
  series: '2hrs ph3'
  model: Arima(order=c(1, 0, 0), seasonal=c(1, 0, 0), method="ML")
  transformation: identity() #log() %>% (function(d) {d[d==-Inf]<-0; d})
  traindays: 7
  testdays: 1
  xreg: NULL
  obs: FALSE
  serial: FALSE
title: '`r params$model` on full `r params$series`'
author: 'Barbu Paul - Gheorghe'
output:
  html_document:
    toc: true
    toc_float: true
    number_sections: true
    self_contained: true
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = FALSE)
```

```{r includes, warning=FALSE, include=FALSE}
library(forecast)
library(ggplot2)
library(tictoc)

source('datasets.R')
source('funcs.R')
```


# Parameters

Series: **`r params$series`**.

Model: **`r params$model`**.

Transformation: **`r params$transformation`**.

As observations: **`r params$obs`**.

Train days: **`r params$traindays`**.

Test days: **`r params$testdays`**.

xreg: **`r params$xreg`**.

Parallel processing: **`r !params$serial`**.

# Original data

The data has been previously cleaned, negative values were made 0.

```{r data, echo=FALSE, message=FALSE, warning=FALSE}
fulldata <- datasets[[params$series]]$series
if(params$obs)
{
  totaldays <- length(fulldata)
}else{
  totaldays <- round(length(fulldata)/frequency(fulldata))
}

fulldata.transformed <- NULL
use_transformation <- !is.null(params$transformation) && params$transformation != 'identity()'
if(use_transformation)
{
  eval(parse(text=paste('fulldata %>%',
                        params$transformation))) -> fulldata.transformed
}else{
  fulldata.transformed <- fulldata
}
```

```{r plot-data, echo=FALSE, message=FALSE, warning=FALSE}
plt <- autoplot(fulldata, series="Original") +
  xlab("Observation [days]") +
  ylab("Energy [Watts]") +
  ggtitle("Original data") + 
  guides(colour=guide_legend(title="Data series"))

if(use_transformation)
{
  plt + autolayer(fulldata.transformed, series="Transformed")
}else{
  plt
}
```

Total data points: **`r length(fulldata)`** representing
**`r totaldays`** days.

Number of data points per day: **`r frequency(fulldata)`**
(gathered once every **`r (24*60)/frequency(fulldata)`** minutes).

# Forecast data

```{r get-forecast, message=FALSE, include=FALSE}
tic('Forecast')

fcastfunc <- NULL

if(params$obs)
{
  if(params$serial)
  {
    fcastfunc <- fullforecast.serial.obs
  }else{
    fcastfunc <- fullforecast.obs
  }
}else{
  if(params$serial)
  {
    fcastfunc <- fullforecast.serial
  }else{
    fcastfunc <- fullforecast  
  }
}

fcast <- fcastfunc(fulldata,
                  params$transformation, 
                  params$model, 
                  params$traindays,
                  params$testdays,
                  params$xreg)
fcast.times <- toc()
```

Time elapapsed for forecasting
**`r totaldays - params$traindays`**
`r ifelse(params$obs, "data points", paste("days (representing", frequency(fulldata) * (totaldays - params$traindays), "data points)"))`,
initial training data not taken into account:
**`r fcast.times$toc - fcast.times$tic`** seconds
(**`r (fcast.times$toc - fcast.times$tic)/60`** minutes).

`r current.rendered.filename.withext <- get(env = parent.frame(n = 8), 'output_file')`

```{r save-forecasts, echo=FALSE}
#get(env = parent.frame(n = 8), 'output_dir')
# current.rendered.filename.withext <- get(env = parent.frame(n = 18), 'output_file')
 current.rendered.filename <- tools::file_path_sans_ext(current.rendered.filename.withext)
 
 save(fcast, file=paste(
   current.rendered.filename,
   ".RData", 
   sep=""))
 
 write.csv(fcast$points, file=paste(
   current.rendered.filename, 
   ".points.csv", 
   sep=""))
 
 write.csv(fcast$adjpoints, file=paste(
   current.rendered.filename, 
   ".adjpoints.csv", 
   sep=""))
```

## Forecasts (un-adjusted) plot against the data
```{r plot-forecast-unadjusted, echo=FALSE}

autoplot(fulldata.transformed,
         series=ifelse(use_transformation, "Transformed", "Original")) +
  autolayer(fcast$points, series="Forecasts") +
  xlab("Observation [days]") +
  ylab("Energy [Watts]") +
  ggtitle("Original data and forecasts") + 
  guides(colour=guide_legend(title="Data series"))
```

## Accuracy of the un-adjusted forecasts against the data
```{r forecast-accuracy, echo=FALSE}
knitr::kable(fcast$accuracy)
```

## Forecasts (adjusted) plot against the data
```{r plot-forecast-adjusted, echo=FALSE}
# autoplot(fulldata.transformed,
#          series=ifelse(use_transformation, "Transformed", "Original")) +
#   autolayer(fcast$adjpoints, series="Forecasts") +
#   xlab("Observation [days]") +
#   ylab("Energy [Watts]") +
#   ggtitle("Original data and forecasts") + 
#   guides(colour=guide_legend(title="Data series"))

lbls <- c("Original", "Forecasts")
autoplot(ts.union(fulldata.transformed, fcast$adjpoints)) +
  geom_point(aes(shape=series)) +
  scale_shape_discrete(labels=lbls) +
  scale_color_manual(labels=lbls, values=c("#00BFC4", "#F8766D")) +
  labs(shape="Data series", color="Data series") +
  xlab("Observation [days]") +
  ylab("Energy [Watts]") +
  ggtitle("Original data and forecasts")
```

## Accuracy of the adjusted forecasts against the data
```{r forecast-adjaccuracy, echo=FALSE}
knitr::kable(fcast$adjaccuracy)
```