Random Forest, using Ranger

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

How it works

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

library(dplyr)
library(tidypredict)
library(ranger)

model <- ranger(mpg ~ ., data = mtcars, num.trees = 5, max.depth = 2)

Under the hood

The parser is based on the output from the ranger::treeInfo() function. It will return as many decision paths as there are non-NA rows in the prediction field.

treeInfo(model) %>%
  head()
#>   nodeID leftChild rightChild splitvarID splitvarName splitval terminal
#> 1      0         1          2          8         gear     3.50    FALSE
#> 2      1         3          4          2           hp   192.50    FALSE
#> 3      2         5          6          4           wt     2.26    FALSE
#> 4      3        NA         NA         NA         <NA>       NA     TRUE
#> 5      4        NA         NA         NA         <NA>       NA     TRUE
#> 6      5        NA         NA         NA         <NA>       NA     TRUE
#>   prediction
#> 1         NA
#> 2         NA
#> 3         NA
#> 4   16.02000
#> 5   12.18333
#> 6   29.98750

The output from parse_model() is transformed into a dplyr, a.k.a. Tidy Eval, formula. Each decision tree becomes one dplyr::case_when() statement, which are then combined.

tidypredict_fit(model)
#> (case_when(is.na(gear) | gear <= 3.5 ~ case_when(is.na(hp) | 
#>     hp <= 192.5 ~ 16.02, .default = 12.1833333333333), .default = case_when(is.na(wt) | 
#>     wt <= 2.26 ~ 29.9875, .default = 20.0076923076923)) + case_when(is.na(wt) | 
#>     wt <= 3.295 ~ case_when(is.na(vs) | vs <= 0.5 ~ 21.1833333333333, 
#>     .default = 25.8714285714286), .default = case_when(is.na(qsec) | 
#>     qsec <= 18.15 ~ 14.1588235294118, .default = 18.5)) + case_when(is.na(disp) | 
#>     disp <= 163.8 ~ case_when(is.na(hp) | hp <= 79.5 ~ 28.125, 
#>     .default = 21.225), .default = case_when(is.na(wt) | wt <= 
#>     4.5475 ~ 17.15, .default = 10.4)) + case_when(is.na(cyl) | 
#>     cyl <= 5 ~ case_when(is.na(disp) | disp <= 101.55 ~ 31.65, 
#>     .default = 23.3), .default = case_when(is.na(cyl) | cyl <= 
#>     7 ~ 20.2666666666667, .default = 15.3538461538462)) + case_when(is.na(cyl) | 
#>     cyl <= 5 ~ case_when(is.na(wt) | wt <= 1.885 ~ 31.5666666666667, 
#>     .default = 23.9714285714286), .default = case_when(is.na(cyl) | 
#>     cyl <= 7 ~ 19.8, .default = 14.91875)))/5

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

parsnip

tidypredict also supports ranger model objects fitted via the parsnip package.

library(parsnip)

parsnip_model <- rand_forest(mode = "regression", trees = 5) %>%
  set_engine("ranger", max.depth = 2) %>%
  fit(mpg ~ ., data = mtcars)

tidypredict_fit(parsnip_model)
#> (case_when(is.na(gear) | gear <= 3.5 ~ case_when(is.na(disp) | 
#>     disp <= 197.95 ~ 21.5, .default = 15.42), .default = case_when(is.na(drat) | 
#>     drat <= 4 ~ 23.4444444444444, .default = 27.9833333333333)) + 
#>     case_when(is.na(hp) | hp <= 131.5 ~ case_when(is.na(wt) | 
#>         wt <= 2.2775 ~ 30.5, .default = 21.1125), .default = case_when(is.na(drat) | 
#>         drat <= 3.035 ~ 10.4, .default = 16.8833333333333)) + 
#>     case_when(is.na(cyl) | cyl <= 5 ~ case_when(is.na(hp) | hp <= 
#>         78.5 ~ 31.15, .default = 26.1285714285714), .default = case_when(is.na(disp) | 
#>         disp <= 266.9 ~ 20.2, .default = 15.2583333333333)) + 
#>     case_when(is.na(vs) | vs <= 0.5 ~ case_when(is.na(drat) | 
#>         drat <= 4.325 ~ 16.7, .default = 26), .default = case_when(is.na(wt) | 
#>         wt <= 2.26 ~ 32.2333333333333, .default = 20.6375)) + 
#>     case_when(is.na(disp) | disp <= 120.65 ~ case_when(is.na(vs) | 
#>         vs <= 0.5 ~ 26, .default = 31.2777777777778), .default = case_when(is.na(wt) | 
#>         wt <= 3.3125 ~ 21.4555555555556, .default = 16.4307692307692)))/5