---
title: "Random Forest - partykit"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Random Forest - partykit}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r pre, include = FALSE}
if (!rlang::is_installed("partykit")) {
  knitr::opts_chunk$set(
    eval = FALSE
  )
}
```

```{r, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(dplyr)
library(tidypredict)
library(partykit)
library(parsnip)
set.seed(100)
```

| Function                                                      |Works|
|---------------------------------------------------------------|-----|
|`tidypredict_fit()`, `tidypredict_sql()`, `parse_model()`      |  ✔  |
|`tidypredict_to_column()`                                      |  ✔  |
|`tidypredict_test()`                                           |  ✔  |
|`tidypredict_interval()`, `tidypredict_sql_interval()`         |  ✗  |
|`parsnip`                                                      |  ✔  |

Only regression models are supported. Classification `cforest()` models rely on
a voting mechanism that cannot be expressed as a single formula.

## How it works

Here is a simple `cforest()` model using the `mtcars` dataset:

```{r}
library(dplyr)
library(tidypredict)
library(partykit)

model <- cforest(mpg ~ wt + cyl, data = mtcars, ntree = 5)
```

## Under the hood

Each tree in the forest is a `partykit` party tree. `tidypredict` turns every
tree into a nested `dplyr::case_when()` statement using each terminal node's
in-bag weighted mean, then averages the trees. This matches the default
`predict()` behavior, which scales the per-tree weights before aggregating.

```{r}
tidypredict_fit(model)
```

From there, the Tidy Eval formula can be used anywhere where it can be operated.
`tidypredict` provides three paths:

  - Use directly inside `dplyr`, `mutate(mtcars, !!tidypredict_fit(model))`
  - Use `tidypredict_to_column(model)` to a piped command set
  - Use `tidypredict_sql(model, con)` to retrieve the SQL statement

## parsnip

`tidypredict` also supports `cforest` model objects fitted via the `parsnip`
package with the `"partykit"` engine, which is provided by the `bonsai` package.

```{r, eval = rlang::is_installed("bonsai")}
library(bonsai)
library(parsnip)

parsnip_model <- rand_forest(mode = "regression", trees = 5) %>%
  set_engine("partykit") %>%
  fit(mpg ~ wt + cyl, data = mtcars)

tidypredict_fit(parsnip_model)
```