03_data_exploration.Rmd

---
title: 'Data exploration'
author: "Janko Thyson"
diesel_he: "14 Mai 2017"
output: 
  html_document:
    toc: true
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
# knitr::opts_knit$set(root.dir = "..")
suppressWarnings({
  library(dygraphs)
  library(xts)
  library(forecast)
  library(ggfortify)
})
```

## Load diesel_ha 

```{r}
diesel_h <- readRDS("data/diesel_h.RDS")
# diesel_h %>% attributes()
# diesel_h %>% head()
names(diesel_h) <- "diesel"
```

## Trend

### Interactive

```{r}
dygraph(diesel_h) %>% dyRoller(rollPeriod = 24)
```

### All in one

```{r}
# Smooth "within day" >> how does a day look like?
ma_24 <- rollmean(diesel_h, 24, fill = list(NA, NULL, NA))

# Smooth "within week" >> how does a week look like?
ma_168 <- rollmean(diesel_h, 24 * 7, fill = list(NA, NULL, NA))

# Smooth "within month" >> how does a (standard) month look like?
ma_5114 <- rollmean(diesel_h, round(24 * 7 * 30.44), fill = list(NA, NULL, NA))

diesel_h_viz <- cbind(diesel_h, ma_24, ma_168, ma_5114)
names(diesel_h_viz) <- c("value", "ma_24", "ma_168", "ma_5114")
dygraph(diesel_h_viz)
```

## Seasonality

### Hour in week 

```{r}
attributes(diesel_h)$frequency <- 24 * 7
```

#### Season plot

```{r}
ggseasonplot(as.ts(diesel_h), year.labels = FALSE)
```

#### Subseason plot

```{r}
ggmonthplot(as.ts(diesel_h))
```

#### NOTE
**JUST AN EXAMPLE**
Apply to correct/suitable aggregation level, doesn't make sense otherwise 

```{r}
ggfreqplot(as.ts(diesel_h), freq = 4)
```

-----

### Hour in year 

```{r}
attributes(diesel_h)$frequency <- 24 * 7 * 52
```

#### Season plot

```{r}
ggseasonplot(as.ts(diesel_h))
```

#### Subseason plot

```{r}
ggmonthplot(as.ts(diesel_h))
```

#### NOTE
**JUST AN EXAMPLE**
Apply to correct/suitable aggregation level, doesn't make sense otherwise 

```{r}
ggfreqplot(as.ts(diesel_h), freq = 4)
```