[英]Tidymodel Package: General linear models (glm) and decision tree (bagged trees, boosted trees, and random forest) models in R
問題
我正在嘗試使用R 中的Tidymodels 包進行分析。 我正在關注以下有關 R 中決策樹學習的教程：-
教程
https://bcullen.rbind.io/post/2020-06-02-tidymodels-decision-tree-learning-in-r/
我有一個名為 FID （見下文）的數據框，其中因變量是頻率（數字） ，預測變量是：-年（數字）、月（因子）、季風（因子）和天（數字）。
我相信我已經通過構建袋裝樹、隨機森林和提升樹模型成功地遵循了名為“Tidymodels：R 中的決策樹學習”的教程。
對於這個分析，我還想構建一個通用線性模型（glm） ，以便在所有模型（即隨機森林、袋裝樹、提升樹和通用線性模型）之間進行模型比較，以建立最佳模型擬合。 所有模型都經過10 倍交叉驗證以減少過度擬合的偏差。
問題
隨后，我嘗試調整教程中的代碼（請參見下文）以適應 glm 模型，但我很困惑我是否適當地調整了模型。 當我嘗試在模型全部擬合后生成rmse值時，我不確定 glm R-code 的這個元素是否會產生以下錯誤消息：-
錯誤信息
Error: Problem with `mutate()` input `model`.
x Input `model` can't be recycled to size 4.
ℹ Input `model` is `c("bag", "rf", "boost")`.
ℹ Input `model` must be size 4 or 1, not 3.
此外，我不確定這些模型的 recipe() 函數中表達的R 代碼是否足夠或正確，這在擬合每個模型之前的處理步驟中非常重要。 從我的角度來看，我想知道模型的配方是否可以改進。
如果這是可能的，我想知道是否有人可以幫助我了解擬合 glm 模型時的錯誤消息，並結合更正配方（如果有必要）。
我不是一個高級的 R 編碼員，我已經通過研究其他 Tidymodel 教程進行了徹底的嘗試，試圖找到解決方案； 但是，我不明白此錯誤消息的含義。 因此，如果有人能夠提供幫助，我要表示最深切的感謝。
提前謝謝了。
代碼
##Open the tidymodels package
library(tidymodels)
library(glmnet)
library(parsnip)
library(rpart.plot)
library(rpart)
library(tidyverse) # manipulating data
library(skimr) # data visualization
library(baguette) # bagged trees
library(future) # parallel processing & decrease computation time
library(xgboost) # boosted trees
library(ranger)
###########################################################
# Put 3/4 of the data into the training set
#split this single dataset into two: a training set and a testing set
data_split <- initial_split(Tidy_df, prop = 3/4)
# Create data frames for the two sets:
train_data <- training(data_split)
test_data <- testing(data_split)
# resample the data with 10-fold cross-validation (10-fold by default)
cv <- vfold_cv(train_data)
###########################################################
##Produce the recipe
##Preprocessing
############################################################
rec <- recipe(Frequency ~ ., data = fid_df) %>%
step_nzv(all_predictors(), freq_cut = 0, unique_cut = 0) %>% # remove variables with zero variances
step_novel(all_nominal()) %>% # prepares test data to handle previously unseen factor levels
step_medianimpute(all_numeric(), -all_outcomes(), -has_role("id vars")) %>% # replaces missing numeric observations with the median
step_dummy(all_nominal(), -has_role("id vars")) # dummy codes categorical variables
###########################################################
##Create Models
###########################################################
##########################################################
##General Linear Models
#########################################################
##glm
mod_glm<-linear_reg(mode="regression",
penalty = 0.1,
mixture = 1) %>%
set_engine("glmnet")
##Create workflow
wflow_glm <- workflow() %>%
add_recipe(rec) %>%
add_model(mod_glm)
##Fit the model
plan(multisession)
fit_glm <- fit_resamples(
wflow_glm,
cv,
metrics = metric_set(rmse, rsq),
control = control_resamples(save_pred = TRUE)
)
##########################################################
##Bagged Trees
##########################################################
#####Bagged Trees
mod_bag <- bag_tree() %>%
set_mode("regression") %>%
set_engine("rpart", times = 10) #10 bootstrap resamples
##Create workflow
wflow_bag <- workflow() %>%
add_recipe(rec) %>%
add_model(mod_bag)
##Fit the model
plan(multisession)
fit_bag <- fit_resamples(
wflow_bag,
cv,
metrics = metric_set(rmse, rsq),
control = control_resamples(save_pred = TRUE)
)
###################################################
##Random forests
###################################################
mod_rf <-rand_forest(trees = 1e3) %>%
set_engine("ranger",
num.threads = parallel::detectCores(),
importance = "permutation",
verbose = TRUE) %>%
set_mode("regression")
##Create Workflow
wflow_rf <- workflow() %>%
add_model(mod_rf) %>%
add_recipe(rec)
##Fit the model
plan(multisession)
fit_rf<-fit_resamples(
wflow_rf,
cv,
metrics = metric_set(rmse, rsq),
control = control_resamples(save_pred = TRUE)
)
############################################################
##Boosted Trees
############################################################
mod_boost <- boost_tree() %>%
set_engine("xgboost", nthreads = parallel::detectCores()) %>%
set_mode("regression")
##Create workflow
wflow_boost <- workflow() %>%
add_recipe(rec) %>%
add_model(mod_boost)
##Fit model
plan(multisession)
fit_boost <-fit_resamples(
wflow_boost,
cv,
metrics = metric_set(rmse, rsq),
control = control_resamples(save_pred = TRUE)
)
##############################################
##Evaluate the models
##############################################
collect_metrics(fit_bag) %>%
bind_rows(collect_metrics(fit_rf)) %>%
bind_rows(collect_metrics(fit_boost)) %>%
bind_rows(collect_metrics(fit_glm)) %>%
dplyr::filter(.metric == "rmse") %>%
dplyr::mutate(model = c("bag", "rf", "boost")) %>%
dplyr::select(model, everything()) %>%
knitr::kable()
####Error message
Error: Problem with `mutate()` input `model`.
x Input `model` can't be recycled to size 4.
ℹ Input `model` is `c("bag", "rf", "boost")`.
ℹ Input `model` must be size 4 or 1, not 3.
Run `rlang::last_error()` to see where the error occurred.
#####################################################
##Out-of-sample performance
#####################################################
# bagged trees
final_fit_bag <- last_fit(
wflow_bag,
split = split)
# random forest
final_fit_rf <- last_fit(
wflow_rf,
split = split)
# boosted trees
final_fit_boost <- last_fit(
wflow_boost,
split = split)
數據幀 - FID
structure(list(Year = c(2015, 2015, 2015, 2015, 2015, 2015, 2015,
2015, 2015, 2015, 2015, 2015, 2016, 2016, 2016, 2016, 2016, 2016,
2016, 2016, 2016, 2016, 2016, 2016, 2017, 2017, 2017, 2017, 2017,
2017, 2017, 2017, 2017, 2017, 2017, 2017), Month = structure(c(1L,
2L, 3L, 4L, 5L, 6L, 7L, 8L, 9L, 10L, 11L, 12L, 1L, 2L, 3L, 4L,
5L, 6L, 7L, 8L, 9L, 10L, 11L, 12L, 1L, 2L, 3L, 4L, 5L, 6L, 7L,
8L, 9L, 10L, 11L, 12L), .Label = c("January", "February", "March",
"April", "May", "June", "July", "August", "September", "October",
"November", "December"), class = "factor"), Monsoon = structure(c(2L,
2L, 1L, 1L, 4L, 4L, 4L, 4L, 4L, 3L, 3L, 2L, 2L, 2L, 1L, 1L, 4L,
4L, 4L, 4L, 4L, 3L, 3L, 2L, 2L, 2L, 1L, 1L, 4L, 4L, 4L, 4L, 4L,
3L, 3L, 2L), .Label = c("First_Inter_Monssoon", "North_Monsoon",
"Second_Inter_Monsoon", "South_Monsson"), class = "factor"),
Frequency = c(36, 28, 39, 46, 5, 0, 0, 22, 10, 15, 8,
33, 33, 29, 31, 23, 8, 9, 7, 40, 41, 41, 30, 30, 44, 37,
41, 42, 20, 0, 7, 27, 35, 27, 43, 38), Days = c(31,
28, 31, 30, 6, 0, 0, 29, 15, 29, 29, 31, 31, 29, 30, 30,
7, 0, 7, 30, 30, 31, 30, 27, 31, 28, 30, 30, 21, 0, 7, 26,
29, 27, 29, 29)), row.names = c(NA, -36L), class = "data.frame")
我相信擬合線性模型的錯誤來自Month和Monsoon彼此之間的關系。 它們完全相關：
library(tidyverse)
fid_df <- structure(list(Year = c(2015, 2015, 2015, 2015, 2015, 2015, 2015,
2015, 2015, 2015, 2015, 2015, 2016, 2016, 2016, 2016, 2016, 2016,
2016, 2016, 2016, 2016, 2016, 2016, 2017, 2017, 2017, 2017, 2017,
2017, 2017, 2017, 2017, 2017, 2017, 2017), Month = structure(c(1L,
2L, 3L, 4L, 5L, 6L, 7L, 8L, 9L, 10L, 11L, 12L, 1L, 2L, 3L, 4L,
5L, 6L, 7L, 8L, 9L, 10L, 11L, 12L, 1L, 2L, 3L, 4L, 5L, 6L, 7L,
8L, 9L, 10L, 11L, 12L), .Label = c("January", "February", "March",
"April", "May", "June", "July", "August", "September", "October",
"November", "December"), class = "factor"), Monsoon = structure(c(2L,
2L, 1L, 1L, 4L, 4L, 4L, 4L, 4L, 3L, 3L, 2L, 2L, 2L, 1L, 1L, 4L,
4L, 4L, 4L, 4L, 3L, 3L, 2L, 2L, 2L, 1L, 1L, 4L, 4L, 4L, 4L, 4L,
3L, 3L, 2L), .Label = c("First_Inter_Monssoon", "North_Monsoon",
"Second_Inter_Monsoon", "South_Monsson"), class = "factor"),
Frequency = c(36, 28, 39, 46, 5, 0, 0, 22, 10, 15, 8,
33, 33, 29, 31, 23, 8, 9, 7, 40, 41, 41, 30, 30, 44, 37,
41, 42, 20, 0, 7, 27, 35, 27, 43, 38), Days = c(31,
28, 31, 30, 6, 0, 0, 29, 15, 29, 29, 31, 31, 29, 30, 30,
7, 0, 7, 30, 30, 31, 30, 27, 31, 28, 30, 30, 21, 0, 7, 26,
29, 27, 29, 29)), row.names = c(NA, -36L), class = "data.frame")
fid_df %>%
count(Month, Monsoon)
#> Month Monsoon n
#> 1 January North_Monsoon 3
#> 2 February North_Monsoon 3
#> 3 March First_Inter_Monssoon 3
#> 4 April First_Inter_Monssoon 3
#> 5 May South_Monsson 3
#> 6 June South_Monsson 3
#> 7 July South_Monsson 3
#> 8 August South_Monsson 3
#> 9 September South_Monsson 3
#> 10 October Second_Inter_Monsoon 3
#> 11 November Second_Inter_Monsoon 3
#> 12 December North_Monsoon 3
如果您在線性模型中使用這樣的變量，則模型無法找到兩組系數的估計值：
lm(Frequency ~ ., data = fid_df) %>% summary()
#>
#> Call:
#> lm(formula = Frequency ~ ., data = fid_df)
#>
#> Residuals:
#> Min 1Q Median 3Q Max
#> -15.0008 -3.9357 0.6564 2.9769 12.7681
#>
#> Coefficients: (3 not defined because of singularities)
#> Estimate Std. Error t value Pr(>|t|)
#> (Intercept) -7286.9226 3443.9292 -2.116 0.0459 *
#> Year 3.6155 1.7104 2.114 0.0461 *
#> MonthFebruary -3.2641 6.6172 -0.493 0.6267
#> MonthMarch 0.1006 6.5125 0.015 0.9878
#> MonthApril 0.4843 6.5213 0.074 0.9415
#> MonthMay -4.0308 11.0472 -0.365 0.7187
#> MonthJune 1.0135 15.5046 0.065 0.9485
#> MonthJuly -2.6910 13.6106 -0.198 0.8451
#> MonthAugust -4.9307 6.6172 -0.745 0.4641
#> MonthSeptember -1.7105 7.1126 -0.240 0.8122
#> MonthOctober -7.6981 6.5685 -1.172 0.2538
#> MonthNovember -8.7484 6.5493 -1.336 0.1953
#> MonthDecember -1.6981 6.5685 -0.259 0.7984
#> MonsoonNorth_Monsoon NA NA NA NA
#> MonsoonSecond_Inter_Monsoon NA NA NA NA
#> MonsoonSouth_Monsson NA NA NA NA
#> Days 1.1510 0.4540 2.535 0.0189 *
#> ---
#> Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#>
#> Residual standard error: 7.968 on 22 degrees of freedom
#> Multiple R-squared: 0.8135, Adjusted R-squared: 0.7033
#> F-statistic: 7.381 on 13 and 22 DF, p-value: 2.535e-05
由reprex 包(v0.3.0.9001) 於 2020 年 11 月 18 日創建
由於您擁有此信息，我建議您使用一些領域知識來決定是在模型中使用Month還是Monsoon ，但不要同時使用兩者。
回答 Julia Silge 的建議
#split this single dataset into two: a training set and a testing set
data_split <- initial_split(Tidy_df)
# Create data frames for the two sets:
train_data <- training(data_split)
test_data <- testing(data_split)
# resample the data with 10-fold cross-validation (10-fold by default)
cv <- vfold_cv(train_data)
###########################################################
##Produce the recipe
rec <- recipe(Frequency_Blue ~ ., data = Tidy_df) %>%
step_nzv(all_predictors(), freq_cut = 0, unique_cut = 0) %>% # remove variables with zero variances
step_novel(all_nominal()) %>% # prepares test data to handle previously unseen factor levels
step_medianimpute(all_numeric(), -all_outcomes(), -has_role("id vars")) %>% # replaces missing numeric observations with the median
step_dummy(all_nominal(), -has_role("id vars")) # dummy codes categorical variables
###########################################################
##Create Models
###########################################################
##########################################################
##General Linear Models
#########################################################
##glm
mod_glm<-linear_reg(mode="regression",
penalty = 0.1,
mixture = 1) %>%
set_engine("glmnet")
##Create workflow
wflow_glm <- workflow() %>%
add_recipe(rec) %>%
add_model(mod_glm)
##Fit the model
plan(multisession)
fit_glm <- fit_resamples(
wflow_glm,
cv,
metrics = metric_set(rmse, rsq),
control = control_resamples(save_pred = TRUE)
)
##########################################################
##Bagged Trees
##########################################################
#####Bagged Trees
mod_bag <- bag_tree() %>%
set_mode("regression") %>%
set_engine("rpart", times = 10) #10 bootstrap resamples
##Create workflow
wflow_bag <- workflow() %>%
add_recipe(rec) %>%
add_model(mod_bag)
##Fit the model
plan(multisession)
fit_bag <- fit_resamples(
wflow_bag,
cv,
metrics = metric_set(rmse, rsq),
control = control_resamples(save_pred = TRUE)
)
###################################################
##Random forests
###################################################
mod_rf <-rand_forest(trees = 1e3) %>%
set_engine("ranger",
num.threads = parallel::detectCores(),
importance = "permutation",
verbose = TRUE) %>%
set_mode("regression")
##Create Workflow
wflow_rf <- workflow() %>%
add_model(mod_rf) %>%
add_recipe(rec)
##Fit the model
plan(multisession)
fit_rf<-fit_resamples(
wflow_rf,
cv,
metrics = metric_set(rmse, rsq),
control = control_resamples(save_pred = TRUE)
)
############################################################
##Boosted Trees
############################################################
mod_boost <- boost_tree() %>%
set_engine("xgboost", nthreads = parallel::detectCores()) %>%
set_mode("regression")
##Create workflow
wflow_boost <- workflow() %>%
add_recipe(rec) %>%
add_model(mod_boost)
##Fit model
plan(multisession)
fit_boost <-fit_resamples(
wflow_boost,
cv,
metrics = metric_set(rmse, rsq),
control = control_resamples(save_pred = TRUE)
)
##############################################
##Evaluate the models
##############################################
collect_metrics(fit_bag) %>%
bind_rows(collect_metrics(fit_rf)) %>%
bind_rows(collect_metrics(fit_boost)) %>%
bind_rows(collect_metrics(fit_glm)) %>%
dplyr::filter(.metric == "rmse") %>%
dplyr::mutate(model = c("bag", "rf", "boost", "glm")) %>%
dplyr::select(model, everything()) %>%
knitr::kable()
##rmse values for all models
|model |.metric |.estimator | mean| n| std_err|
|:-----|:-------|:----------|---------:|--:|--------:|
|bag |rmse |standard | 8.929936| 10| 1.544587|
|rf |rmse |standard | 10.188710| 10| 1.144354|
|boost |rmse |standard | 9.249904| 10| 1.489482|
|glm |rmse |standard | 11.348420| 10| 2.217807|
#####################################################
##Out-of-sample performance
#####################################################
#glm
# bagged trees
final_fit_glm <- last_fit(
wflow_glm,
split = split)
# bagged trees
final_fit_bag <- last_fit(
wflow_bag,
split = split)
# random forest
final_fit_rf <- last_fit(
wflow_rf,
split = split)
# boosted trees
final_fit_boost <- last_fit(
wflow_boost,
split = split)
R：Tidymodels：是否可以在整潔的模型中使用 plot 隨機森林 model 的樹木？
[英]R: Tidymodels: Is it possible to plot the trees for a random forest model in tidy models?
MLR - 計算袋裝增強樹的特征重要性（XGBoost）
[英]MLR - calculating feature importance for bagged, boosted trees (XGBoost)
將隨機森林變成決策樹-在R中使用randomForest包
[英]Turning a Random Forest into a Decision Tree - Using randomForest package in R
將來自兩個隨機森林模型的樹模型合並到R中H2O的一個隨機森林模型中
[英]Merging Tree Models from two random forest models into one random forest model at H2O in R
在R中的隨機森林樹的每個最終節點處提取類分布
[英]Extracting class distribution at each final node of trees in random forest in R
聲明:本站的技術帖子網頁，遵循CC BY-SA 4.0協議，如果您需要轉載，請注明本站網址或者原文地址。任何問題請咨詢:yoyou2525@163.com.