wrangling.rmd

---
title: Wrangling event data
output: 
  html_document:
    toc: true
    toc_float: 
      collapsed: false
      print: false
---

```{r include = F}
library(bupaR)
```

```{r echo = F}
htmltools::includeHTML("tracking_google_analytics.html")
```

Next to the specifically supported preprocessing steps (aggregating, subsetting and enriching), event logs can also be manipulated in a more generic way. In order to do so, well-known [dplyr](https://dplyr.tidyverse.org/)-verbs have been adapted to event log objects and other extensions have been made. 


## group_by

Using the `group_by` function, event logs can be grouped according to a (set of) variables, such that all further computations happen for each of these different groups.

In the next example, the number of cases are computed for each value of "vehicleclass".

```{r}
traffic_fines %>%
	group_by(vehicleclass) %>%
	n_cases()
```

### Predefined grouping functions

For specific groupings, some auxiliary functions are available. 

*	`group_by_case` - group by cases
*	`group_by_activity` - group by activity types
*	`group_by_resource` - group by resources
*	`group_by_activity_resource` - group by activity resource pair
*	`group_by_activity_instance` - group by activity instances.

For example, the number of cases in which a specific resource occurs, can be computed as follows.

```{r}
sepsis %>%
	group_by_resource %>%
	n_cases
```


Note that each of the descriptive metrics discussed [here](http://www.bupar.net/exploring.html) can be rewritten using these lower-level functions. The example above is equal to the `resource_involvement` metric at case level. 

### Remove grouping 

When a grouping is no longer needed, it can be removed using the `ungroup_eventlog` function. 

## mutate

You can use `mutate` to add new variable to an event log, possibly by using existing variables. In the next example, the total amount of lacticacid is computed for each case. 

```{r echo = F}
sepsis %>%
	mutate(lacticacid = as.numeric(lacticacid)) -> sepsis
```


```{r}
sepsis %>%
	group_by_case() %>%
	mutate(total_lacticacid = sum(lacticacid, na.rm = T))
```

## filter

Generic filtering of events can be done using the `filter` function, which takes an event log and any number of logical conditions. The example below filters events which have vehicleclas "C" and amount greater than 300. More process-specific filtering methods can be found [here](http://www.bupar.net/subsetting.html).

```{r}
traffic_fines %>%
	filter(vehicleclass == "C", amount > 300)
```

## select

Variables on a event log can be _selected_ using `select`. By default, `select` will always make sure that the mapping-variables are retained. Otherwise, it would no longer function as an eventlog.

```{r}
traffic_fines %>%
	select(vehicleclass)
```



By setting the argument `force_df = TRUE`, the mapping-variables will not be retained, and the output will be a data.frame, and not an event log.

```{r}
traffic_fines %>%
	select(case_id, vehicleclass, amount, force_df = TRUE)
```


## arrange

Event data can be sorted using the `arrange` function. `desc` can be used to sort descendingly on an attribute.

```{r}
#sort descending on time
patients %>%
	arrange(desc(time))
```


## slice 

An eventlog can be _sliced_, which mean returning a slice, i.e. a subset, from the eventlog, based on row number. There are three ways to slice event logs

*	Using `slice`: take a slice of cases
*	Using `slice_activities`: take a slice of activity instances
*	Using `slice_events`: take a slice of events

The next piece of code returns the _first_ 10 cases. Note that first here is defined by the current order of the data set, not by time.

```{r}
patients %>%
	slice(1:10)
```

### slice_activities

The next piece of code returns the _first_ 10 activity instances.

```{r}
patients %>%
	slice_activities(1:10)
```


### slice_events

The next piece of code returns the _first_ 10 events. 

```{r}
patients %>% 
	slice_events(1:10)
```


## first_n, last_n

The slice function select events, cases or activity instances based on their current position in the event data. As such, the result can be changed using the `arrange` function. More often, we want to select the first _n_ activity instances, or the last ones. This is achieved with the `first_n` or `last_n` functions, which return the first, resp. last, n activity instances of a log based on time, not on position.

```{r}
patients %>% 
	first_n(n = 5)
```

This is not impacted by a different ordering of the data since it will take the time aspect into account. 

```{r}
patients %>%
	arrange(desc(time)) %>%
	first_n(n = 5)
```

Incombination with `group_by_case`, it is very easy to select the heads or tails of each case. Below, we explore the 95% most common first 3 activities in the `sepsis` log.

```{r}
sepsis %>%
	group_by_case() %>%
	first_n(3) %>%
	trace_explorer(cov = 0.95)
```



## sample_n

The `sample_n` function allows to take a sample of the event log containing n cases. The code below returns a sample of 10 patients.

```{r}
patients %>%
	sample_n(size = 10)
```
Note that this function can also be used with a sample size bigger than the number of cases in the event log, if you allow for the replacements of drawn cases.

A more extensive list of subsetting methods is provided by edeaR. Look [here](http://www.bupar.net/subsetting.html) for more information.