Functions like last_fit() from the tune package produces last_fit objects which are large nested lists containing the fit results. I tried to transform them into data.frames using the tidy() function from the broom package but this resulted in the following error:
MRE :
library(tidymodels)
library(tidyverse)
data <- mtcars
model_default<-
parsnip::boost_tree(
mode = "regression"
) %>%
set_engine('xgboost',objective = 'reg:squarederror')
wf <- workflow() %>%
add_model(model_default) %>%
add_recipe(recipe(mpg~.,data))
lf <- last_fit(wf,split)
tidy_lf <- tidy(lf)
Error in var(if (is.vector(x) || is.factor(x)) x else as.double(x), na.rm = na.rm) :
is.atomic(x) is not TRUE
In addition: Warning messages:
1: Data frame tidiers are deprecated and will be removed in an upcoming release of broom.
2: In mean.default(X[[i]], ...) :
argument is not numeric or logical: returning NA
3: In mean.default(X[[i]], ...) :
argument is not numeric or logical: returning NA
4: In mean.default(X[[i]], ...) :
argument is not numeric or logical: returning NA
5: In mean.default(X[[i]], ...) :
argument is not numeric or logical: returning NA
6: In mean.default(X[[i]], ...) :
argument is not numeric or logical: returning NA
7: In mean.default(X[[i]], ...) :
argument is not numeric or logical: returning NA
Question : How can I use tidy() with an last_fit() output?
The object that last_fit() creates is a tibble (containing metrics, predictions, etc), not a model that can be tidied. You can use extract_workflow() to extract out the fitted workflow from the object created by last_fit(), and this object can be tidied:
library(tidymodels)
car_split <- initial_split(mtcars)
wf <- workflow() %>%
add_model(linear_reg()) %>%
add_recipe(recipe(mpg ~ ., mtcars))
lf <- last_fit(wf, car_split)
lf
#> # Resampling results
#> # Manual resampling
#> # A tibble: 1 × 6
#> splits id .metrics .notes .predictions .workflow
#> <list> <chr> <list> <list> <list> <list>
#> 1 <split [24/8]> train/test split <tibble> <tibble> <tibble [8 × 4]> <workflow>
lf %>%
extract_workflow() %>%
tidy()
#> # A tibble: 11 × 5
#> term estimate std.error statistic p.value
#> <chr> <dbl> <dbl> <dbl> <dbl>
#> 1 (Intercept) -33.6 36.0 -0.935 0.367
#> 2 cyl -0.0296 1.34 -0.0221 0.983
#> 3 disp 0.0252 0.0269 0.934 0.367
#> 4 hp -0.00539 0.0319 -0.169 0.868
#> 5 drat -0.167 2.54 -0.0659 0.948
#> 6 wt -5.69 2.79 -2.04 0.0623
#> 7 qsec 3.32 1.76 1.89 0.0820
#> 8 vs -4.40 3.80 -1.16 0.268
#> 9 am 2.54 2.67 0.950 0.360
#> 10 gear 2.69 2.28 1.18 0.259
#> 11 carb -0.0486 1.11 -0.0439 0.966
Created on 2022-03-23 by the reprex package (v2.0.1)