當(dāng)前位置：首頁 > 编程资源 > 编程问答 >内容正文

编程问答

tidymodels搞定二分类资料多个模型评价和比较

發(fā)布時(shí)間：2023/12/14 编程问答 22 豆豆

生活随笔收集整理的這篇文章主要介紹了 tidymodels搞定二分类资料多个模型评价和比较小編覺得挺不錯(cuò)的,現(xiàn)在分享給大家,幫大家做個(gè)參考.

本文首發(fā)于公眾號(hào)：醫(yī)學(xué)和生信筆記

“

醫(yī)學(xué)和生信筆記，專注R語言在臨床醫(yī)學(xué)中的使用，R語言數(shù)據(jù)分析和可視化。主要分享R語言做醫(yī)學(xué)統(tǒng)計(jì)學(xué)、meta分析、網(wǎng)絡(luò)藥理學(xué)、臨床預(yù)測模型、機(jī)器學(xué)習(xí)、生物信息學(xué)等。

前面介紹了很多二分類資料的模型評(píng)價(jià)內(nèi)容，用到了很多R包，雖然達(dá)到了目的，但是內(nèi)容太多了，不太容易記住。

今天給大家介紹一個(gè)很厲害的R包：tidymodels，一個(gè)R包搞定二分類資料的模型評(píng)價(jià)和比較。

一看這個(gè)名字就知道，和tidyverse系列師出同門，包的作者是大佬Max Kuhn，大佬的上一個(gè)作品是caret，現(xiàn)在加盟rstudio了，開發(fā)了新的機(jī)器學(xué)習(xí)R包，也就是今天要介紹的tidymodels。

給大家看看如何用優(yōu)雅的方式建立、評(píng)價(jià)、比較多個(gè)模型！

本期目錄：

加載數(shù)據(jù)和R包
數(shù)據(jù)劃分
數(shù)據(jù)預(yù)處理
建立多個(gè)模型
- logistic
- knn
- 隨機(jī)森林
- 決策樹
交叉驗(yàn)證
ROC曲線畫一起

加載數(shù)據(jù)和R包

沒有安裝的R包的自己安裝下~

suppressPackageStartupMessages(library(tidyverse))
suppressPackageStartupMessages(library(tidymodels))
tidymodels_prefer()

由于要做演示用，肯定要一份比較好的數(shù)據(jù)才能說明問題，今天用的這份數(shù)據(jù)，結(jié)果變量是一個(gè)二分類的。

一共有91976行，26列，其中play_type是結(jié)果變量，因子型，其余列都是預(yù)測變量。

all_plays?<-?read_rds("../000files/all_plays.rds")
glimpse(all_plays)
##?Rows:?91,976
##?Columns:?26
##?$?game_id????????????????????<dbl>?2017090700,?2017090700,?2017090700,?2017090…
##?$?posteam????????????????????<chr>?"NE",?"NE",?"NE",?"NE",?"NE",?"NE",?"NE",?"…
##?$?play_type??????????????????<fct>?pass,?pass,?run,?run,?pass,?run,?pass,?pass…
##?$?yards_gained???????????????<dbl>?0,?8,?8,?3,?19,?5,?16,?0,?2,?7,?0,?3,?10,?0…
##?$?ydstogo????????????????????<dbl>?10,?10,?2,?10,?7,?10,?5,?2,?2,?10,?10,?10,?…
##?$?down???????????????????????<ord>?1,?2,?3,?1,?2,?1,?2,?1,?2,?1,?1,?2,?3,?1,?2…
##?$?game_seconds_remaining?????<dbl>?3595,?3589,?3554,?3532,?3506,?3482,?3455,?3…
##?$?yardline_100???????????????<dbl>?73,?73,?65,?57,?54,?35,?30,?2,?2,?75,?32,?3…
##?$?qtr????????????????????????<ord>?1,?1,?1,?1,?1,?1,?1,?1,?1,?1,?1,?1,?1,?1,?1…
##?$?posteam_score??????????????<dbl>?0,?0,?0,?0,?0,?0,?0,?0,?0,?0,?7,?7,?7,?7,?7…
##?$?defteam????????????????????<chr>?"KC",?"KC",?"KC",?"KC",?"KC",?"KC",?"KC",?"…
##?$?defteam_score??????????????<dbl>?0,?0,?0,?0,?0,?0,?0,?0,?0,?7,?0,?0,?0,?0,?0…
##?$?score_differential?????????<dbl>?0,?0,?0,?0,?0,?0,?0,?0,?0,?-7,?7,?7,?7,?7,?…
##?$?shotgun????????????????????<fct>?0,?0,?1,?1,?1,?0,?1,?0,?0,?1,?1,?1,?1,?1,?0…
##?$?no_huddle??????????????????<fct>?0,?0,?0,?1,?1,?1,?1,?0,?0,?0,?0,?0,?0,?0,?0…
##?$?posteam_timeouts_remaining?<fct>?3,?3,?3,?3,?3,?3,?3,?3,?3,?3,?3,?3,?3,?3,?3…
##?$?defteam_timeouts_remaining?<fct>?3,?3,?3,?3,?3,?3,?3,?3,?3,?3,?3,?3,?3,?3,?3…
##?$?wp?????????????????????????<dbl>?0.5060180,?0.4840546,?0.5100098,?0.5529816,…
##?$?goal_to_go?????????????????<fct>?0,?0,?0,?0,?0,?0,?0,?1,?1,?0,?0,?0,?0,?0,?0…
##?$?half_seconds_remaining?????<dbl>?1795,?1789,?1754,?1732,?1706,?1682,?1655,?1…
##?$?total_runs?????????????????<dbl>?0,?0,?0,?1,?2,?2,?3,?3,?3,?0,?4,?4,?4,?5,?5…
##?$?total_pass?????????????????<dbl>?0,?1,?2,?2,?2,?3,?3,?4,?5,?0,?5,?6,?7,?7,?8…
##?$?previous_play??????????????<fct>?First?play?of?Drive,?pass,?pass,?run,?run,?…
##?$?in_red_zone????????????????<fct>?0,?0,?0,?0,?0,?0,?0,?1,?1,?0,?0,?0,?0,?1,?1…
##?$?in_fg_range????????????????<fct>?0,?0,?0,?0,?0,?1,?1,?1,?1,?0,?1,?1,?1,?1,?1…
##?$?two_min_drill??????????????<fct>?0,?0,?0,?0,?0,?0,?0,?0,?0,?0,?0,?0,?0,?0,?0…

數(shù)據(jù)劃分

把75%的數(shù)據(jù)用于訓(xùn)練集，剩下的做測試集。

set.seed(20220520)

#?數(shù)據(jù)劃分，根據(jù)play_type分層
split_pbp?<-?initial_split(all_plays,?0.75,?strata?=?play_type)

train_data?<-?training(split_pbp)?#?訓(xùn)練集
test_data?<-?testing(split_pbp)?#?測試集

數(shù)據(jù)預(yù)處理

pbp_rec?<-?recipe(play_type?~?.,?data?=?train_data)??%>%
??step_rm(half_seconds_remaining,yards_gained,?game_id)?%>%?#?移除這3列
??step_string2factor(posteam,?defteam)?%>%??#?變?yōu)橐蜃宇愋?/span>
??#update_role(yards_gained,?game_id,?new_role?=?"ID")?%>%?
??#?去掉高度相關(guān)的變量
??step_corr(all_numeric(),?threshold?=?0.7)?%>%?
??step_center(all_numeric())?%>%??#?中心化
??step_zv(all_predictors())??#?去掉零方差變量

建立多個(gè)模型

logistic

選擇模型，連接數(shù)據(jù)預(yù)處理步驟。

lm_spec?<-?logistic_reg(mode?=?"classification",engine?=?"glm")
lm_wflow?<-?workflow()?%>%?
??add_recipe(pbp_rec)?%>%?
??add_model(lm_spec)

建立模型：

fit_lm?<-?lm_wflow?%>%?fit(data?=?train_data)

應(yīng)用于測試集：

pred_lm?<-?select(test_data,?play_type)?%>%?
??bind_cols(predict(fit_lm,?test_data,?type?=?"prob"))?%>%?
??bind_cols(predict(fit_lm,?test_data))

查看模型表現(xiàn)：

#?選擇多種評(píng)價(jià)指標(biāo)
metricsets?<-?metric_set(accuracy,?mcc,?f_meas,?j_index)

pred_lm?%>%?metricsets(truth?=?play_type,?estimate?=?.pred_class)
##?#?A?tibble:?4?×?3
##???.metric??.estimator?.estimate
##???<chr>????<chr>??????????<dbl>
##?1?accuracy?binary?????????0.724
##?2?mcc??????binary?????????0.423
##?3?f_meas???binary?????????0.774
##?4?j_index??binary?????????0.416

大家最喜歡的AUC：

pred_lm?%>%?roc_auc(truth?=?play_type,?.pred_pass)
##?#?A?tibble:?1?×?3
##???.metric?.estimator?.estimate
##???<chr>???<chr>??????????<dbl>
##?1?roc_auc?binary?????????0.781

可視化結(jié)果，首先是大家喜聞樂見的ROC曲線：

pred_lm?%>%?roc_curve(truth?=?play_type,?.pred_pass)?%>%?
??autoplot()
plot of chunk unnamed-chunk-10

pr曲線：

pred_lm?%>%?pr_curve(truth?=?play_type,?.pred_pass)?%>%?
??autoplot()
plot of chunk unnamed-chunk-11

gain_curve：

pred_lm?%>%?gain_curve(truth?=?play_type,?.pred_pass)?%>%?
??autoplot()
plot of chunk unnamed-chunk-12

lift_curve：

pred_lm?%>%?lift_curve(truth?=?play_type,?.pred_pass)?%>%?
??autoplot()
plot of chunk unnamed-chunk-13

混淆矩陣：

pred_lm?%>%?
??conf_mat(play_type,.pred_class)?%>%?
??autoplot()
plot of chunk unnamed-chunk-14

knn

k最近鄰法，和上面的邏輯回歸一模一樣的流程。

首先也是選擇模型，連接數(shù)據(jù)預(yù)處理步驟：

knn_spec?<-?nearest_neighbor(mode?=?"classification",?engine?=?"kknn")

knn_wflow?<-?workflow()?%>%?
??add_recipe(pbp_rec)?%>%?
??add_model(knn_spec)

建立模型：

library(kknn)
fit_knn?<-?knn_wflow?%>%?
??fit(train_data)

應(yīng)用于測試集：

pred_knn?<-?test_data?%>%?select(play_type)?%>%?
??bind_cols(predict(fit_knn,?test_data,?type?=?"prob"))?%>%?
??bind_cols(predict(fit_knn,?test_data,?type?=?"class"))

查看模型表現(xiàn)：

metricsets?<-?metric_set(accuracy,?mcc,?f_meas,?j_index)

pred_knn?%>%?metricsets(truth?=?play_type,?estimate?=?.pred_class)
##?#?A?tibble:?4?×?3
##???.metric??.estimator?.estimate
##???<chr>????<chr>??????????<dbl>
##?1?accuracy?binary?????????0.672
##?2?mcc??????binary?????????0.317
##?3?f_meas???binary?????????0.727
##?4?j_index??binary?????????0.315
pred_knn?%>%?roc_auc(play_type,?.pred_pass)
##?#?A?tibble:?1?×?3
##???.metric?.estimator?.estimate
##???<chr>???<chr>??????????<dbl>
##?1?roc_auc?binary?????????0.718

可視化模型的部分就不說了，和上面的一模一樣！

隨機(jī)森林

同樣的流程來第3遍！

rf_spec?<-?rand_forest(mode?=?"classification")?%>%?
??set_engine("ranger",importance?=?"permutation")
rf_wflow?<-?workflow()?%>%?
??add_recipe(pbp_rec)?%>%?
??add_model(rf_spec)

建立模型：

fit_rf?<-?rf_wflow?%>%?
??fit(train_data)

應(yīng)用于測試集：

pred_rf?<-?test_data?%>%?select(play_type)?%>%?
??bind_cols(predict(fit_rf,?test_data,?type?=?"prob"))?%>%?
??bind_cols(predict(fit_rf,?test_data,?type?=?"class"))

查看模型表現(xiàn)：

pred_rf?%>%?metricsets(truth?=?play_type,?estimate?=?.pred_class)
##?#?A?tibble:?4?×?3
##???.metric??.estimator?.estimate
##???<chr>????<chr>??????????<dbl>
##?1?accuracy?binary?????????0.731
##?2?mcc??????binary?????????0.441
##?3?f_meas???binary?????????0.774
##?4?j_index??binary?????????0.439
pred_rf?%>%?conf_mat(truth?=?play_type,?estimate?=?.pred_class)
##???????????Truth
##?Prediction??pass???run
##???????pass?10622??3225
##???????run???2962??6186
pred_rf?%>%?roc_auc(play_type,?.pred_pass)
##?#?A?tibble:?1?×?3
##???.metric?.estimator?.estimate
##???<chr>???<chr>??????????<dbl>
##?1?roc_auc?binary?????????0.799

下面給大家手動(dòng)畫一個(gè)校準(zhǔn)曲線。

兩種畫法，差別不大，主要是分組方法不一樣，第2種分組方法是大家常見的哦~

calibration_df?<-?pred_rf?%>%?
???mutate(pass?=?if_else(play_type?==?"pass",?1,?0),
??????????pred_rnd?=?round(.pred_pass,?2)
??????????)?%>%?
??group_by(pred_rnd)?%>%?
??summarize(mean_pred?=?mean(.pred_pass),
????????????mean_obs?=?mean(pass),
????????????n?=?n()
????????????)

ggplot(calibration_df,?aes(mean_pred,?mean_obs))+?
??geom_point(aes(size?=?n),?alpha?=?0.5)+
??geom_abline(linetype?=?"dashed")+
??theme_minimal()
plot of chunk unnamed-chunk-26

第2種方法：

cali_df?<-?pred_rf?%>%?
??arrange(.pred_pass)?%>%?
??mutate(pass?=?if_else(play_type?==?"pass",?1,?0),
?????????group?=?c(rep(1:249,each=92),?rep(250,87))
?????????)?%>%?
??group_by(group)?%>%?
??summarise(mean_pred?=?mean(.pred_pass),
????????????mean_obs?=?mean(pass)
????????????)

cali_plot?<-?ggplot(cali_df,?aes(mean_pred,?mean_obs))+?
??geom_point(alpha?=?0.5)+
??geom_abline(linetype?=?"dashed")+
??theme_minimal()

cali_plot
plot of chunk unnamed-chunk-27

隨機(jī)森林這種方法是可以計(jì)算變量重要性的，當(dāng)然也是能把結(jié)果可視化的。

給大家演示下如何可視化隨機(jī)森林結(jié)果的變量重要性：

library(vip)

fit_rf?%>%?
??extract_fit_parsnip()?%>%?
??vip(num_features?=?10)
plot of chunk unnamed-chunk-28

決策樹

同樣的流程來第4遍！不知道你看懂了沒有。。。

tree_spec?<-?decision_tree(mode?=?"classification",engine?=?"rpart")
tree_wflow?<-?workflow()?%>%?
??add_recipe(pbp_rec)?%>%?
??add_model(tree_spec)

建立模型：

fit_tree?<-?tree_wflow?%>%?
??fit(train_data)

應(yīng)用于測試集：

pred_tree?<-?test_data?%>%?select(play_type)?%>%?
??bind_cols(predict(fit_tree,?test_data,?type?=?"prob"))?%>%?
??bind_cols(predict(fit_tree,?test_data,?type?=?"class"))

查看結(jié)果：

pred_tree?%>%?roc_auc(play_type,?.pred_pass)
##?#?A?tibble:?1?×?3
##???.metric?.estimator?.estimate
##???<chr>???<chr>??????????<dbl>
##?1?roc_auc?binary?????????0.706
pred_tree?%>%?metricsets(truth?=?play_type,?estimate?=?.pred_class)
##?#?A?tibble:?4?×?3
##???.metric??.estimator?.estimate
##???<chr>????<chr>??????????<dbl>
##?1?accuracy?binary?????????0.721
##?2?mcc??????binary?????????0.417
##?3?f_meas???binary?????????0.770
##?4?j_index??binary?????????0.411

交叉驗(yàn)證

交叉驗(yàn)證也是大家喜聞樂見的，就用隨機(jī)森林給大家順便演示下交叉驗(yàn)證。

首先要選擇重抽樣方法，這里我們選擇10折交叉驗(yàn)證：

set.seed(20220520)

folds?<-?vfold_cv(train_data,?v?=?10)
folds
##?#??10-fold?cross-validation?
##?#?A?tibble:?10?×?2
##????splits???????????????id????
##????<list>???????????????<chr>?
##??1?<split?[62082/6899]>?Fold01
##??2?<split?[62083/6898]>?Fold02
##??3?<split?[62083/6898]>?Fold03
##??4?<split?[62083/6898]>?Fold04
##??5?<split?[62083/6898]>?Fold05
##??6?<split?[62083/6898]>?Fold06
##??7?<split?[62083/6898]>?Fold07
##??8?<split?[62083/6898]>?Fold08
##??9?<split?[62083/6898]>?Fold09
##?10?<split?[62083/6898]>?Fold10

然后就是讓模型在訓(xùn)練集上跑起來：

keep_pred?<-?control_resamples(save_pred?=?T,?verbose?=?T)

set.seed(20220520)

library(doParallel)?
##?Loading?required?package:?foreach
##?
##?Attaching?package:?'foreach'
##?The?following?objects?are?masked?from?'package:purrr':
##?
##?????accumulate,?when
##?Loading?required?package:?iterators
##?Loading?required?package:?parallel

cl?<-?makePSOCKcluster(12)?#?加速，用12個(gè)線程
registerDoParallel(cl)

rf_res?<-?fit_resamples(rf_wflow,?resamples?=?folds,?control?=?keep_pred)

i?Fold01:?preprocessor?1/1
??Fold01:?preprocessor?1/1
i?Fold01:?preprocessor?1/1,?model?1/1
??Fold01:?preprocessor?1/1,?model?1/1
i?Fold01:?preprocessor?1/1,?model?1/1?(predictions)
i?Fold02:?preprocessor?1/1
??Fold02:?preprocessor?1/1
i?Fold02:?preprocessor?1/1,?model?1/1
??Fold02:?preprocessor?1/1,?model?1/1
i?Fold02:?preprocessor?1/1,?model?1/1?(predictions)
i?Fold03:?preprocessor?1/1
??Fold03:?preprocessor?1/1
i?Fold03:?preprocessor?1/1,?model?1/1
??Fold03:?preprocessor?1/1,?model?1/1
i?Fold03:?preprocessor?1/1,?model?1/1?(predictions)
i?Fold04:?preprocessor?1/1
??Fold04:?preprocessor?1/1
i?Fold04:?preprocessor?1/1,?model?1/1
??Fold04:?preprocessor?1/1,?model?1/1
i?Fold04:?preprocessor?1/1,?model?1/1?(predictions)
i?Fold05:?preprocessor?1/1
??Fold05:?preprocessor?1/1
i?Fold05:?preprocessor?1/1,?model?1/1
??Fold05:?preprocessor?1/1,?model?1/1
i?Fold05:?preprocessor?1/1,?model?1/1?(predictions)
i?Fold06:?preprocessor?1/1
??Fold06:?preprocessor?1/1
i?Fold06:?preprocessor?1/1,?model?1/1
??Fold06:?preprocessor?1/1,?model?1/1
i?Fold06:?preprocessor?1/1,?model?1/1?(predictions)
i?Fold07:?preprocessor?1/1
??Fold07:?preprocessor?1/1
i?Fold07:?preprocessor?1/1,?model?1/1
??Fold07:?preprocessor?1/1,?model?1/1
i?Fold07:?preprocessor?1/1,?model?1/1?(predictions)
i?Fold08:?preprocessor?1/1
??Fold08:?preprocessor?1/1
i?Fold08:?preprocessor?1/1,?model?1/1
??Fold08:?preprocessor?1/1,?model?1/1
i?Fold08:?preprocessor?1/1,?model?1/1?(predictions)
i?Fold09:?preprocessor?1/1
??Fold09:?preprocessor?1/1
i?Fold09:?preprocessor?1/1,?model?1/1
??Fold09:?preprocessor?1/1,?model?1/1
i?Fold09:?preprocessor?1/1,?model?1/1?(predictions)
i?Fold10:?preprocessor?1/1
??Fold10:?preprocessor?1/1
i?Fold10:?preprocessor?1/1,?model?1/1
??Fold10:?preprocessor?1/1,?model?1/1
i?Fold10:?preprocessor?1/1,?model?1/1?(predictions)

stopCluster(cl)

查看模型表現(xiàn)：

rf_res?%>%?
??collect_metrics(summarize?=?T)
##?#?A?tibble:?2?×?6
##???.metric??.estimator??mean?????n?std_err?.config?????????????
##???<chr>????<chr>??????<dbl>?<int>???<dbl>?<chr>???????????????
##?1?accuracy?binary?????0.732????10?0.00157?Preprocessor1_Model1
##?2?roc_auc??binary?????0.799????10?0.00193?Preprocessor1_Model1

查看具體的結(jié)果：

rf_res?%>%?collect_predictions()
##?#?A?tibble:?68,981?×?7
##????id?????.pred_pass?.pred_run??.row?.pred_class?play_type?.config?????????????
##????<chr>???????<dbl>?????<dbl>?<int>?<fct>???????<fct>?????<chr>???????????????
##??1?Fold01??????0.572????0.428??????6?pass????????pass??????Preprocessor1_Model1
##??2?Fold01??????0.470????0.530??????8?run?????????pass??????Preprocessor1_Model1
##??3?Fold01??????0.898????0.102?????22?pass????????pass??????Preprocessor1_Model1
##??4?Fold01??????0.915????0.0847????69?pass????????pass??????Preprocessor1_Model1
##??5?Fold01??????0.841????0.159?????97?pass????????pass??????Preprocessor1_Model1
##??6?Fold01??????0.931????0.0688???112?pass????????pass??????Preprocessor1_Model1
##??7?Fold01??????0.729????0.271????123?pass????????pass??????Preprocessor1_Model1
##??8?Fold01??????0.640????0.360????129?pass????????pass??????Preprocessor1_Model1
##??9?Fold01??????0.740????0.260????136?pass????????pass??????Preprocessor1_Model1
##?10?Fold01??????0.902????0.0979???143?pass????????pass??????Preprocessor1_Model1
##?#?…?with?68,971?more?rows

可視化結(jié)果也是和上面的一模一樣，就不一一介紹了，簡單說下訓(xùn)練集的校準(zhǔn)曲線畫法，其實(shí)也是和上面一樣的~

res_calib_plot?<-?collect_predictions(rf_res)?%>%?
??mutate(
????pass?=?if_else(play_type?==?"pass",?1,?0),
????pred_rnd?=?round(.pred_pass,?2)
????)?%>%?
??group_by(pred_rnd)?%>%
??summarize(
????mean_pred?=?mean(.pred_pass),
????mean_obs?=?mean(pass),
????n?=?n()
????)?%>%?
??ggplot(aes(x?=?mean_pred,?y?=?mean_obs))?+
??geom_abline(linetype?=?"dashed")?+
??geom_point(aes(size?=?n),?alpha?=?0.5)?+
??theme_minimal()?+
??labs(
????x?=?"Predicted?Pass",?
????y?=?"Observed?Pass"
????)?+
??coord_cartesian(
????xlim?=?c(0,1),?ylim?=?c(0,?1)
????)

res_calib_plot
plot of chunk unnamed-chunk-38

然后就是應(yīng)用于測試集，并查看測試集上的表現(xiàn)：

rf_test_res?<-?last_fit(rf_wflow,?split_pbp)?%>%?
??collect_metrics()
##?Error?in?summary.connection(connection):?invalid?connection

rf_test_res
#?A?tibble:?2?×?4
??.metric??.estimator?.estimate?.config?????????????
??<chr>????<chr>??????????<dbl>?<chr>???????????????
1?accuracy?binary?????????0.730?Preprocessor1_Model1
2?roc_auc??binary?????????0.798?Preprocessor1_Model1

ROC曲線畫一起

其實(shí)非常簡單，就是把結(jié)果拼在一起畫個(gè)圖就行了~

roc_lm?<-?pred_lm?%>%?roc_curve(play_type,?.pred_pass)?%>%?
??mutate(model?=?"logistic")

roc_knn?<-?pred_knn?%>%?roc_curve(play_type,?.pred_pass)?%>%?
??mutate(model?=?"kknn")

roc_rf?<-?pred_rf?%>%?roc_curve(play_type,?.pred_pass)?%>%?
??mutate(model?=?"randomforest")

roc_tree?<-?pred_tree?%>%?roc_curve(play_type,?.pred_pass)?%>%?
??mutate(model?=?"decision?tree")

rocs?<-?bind_rows(roc_lm,roc_knn,roc_rf,roc_tree)?%>%?
??ggplot(aes(x?=?1?-?specificity,?y?=?sensitivity,?color?=?model))+
??geom_path(lwd?=?1.2,?alpha?=?0.6)+
??geom_abline(lty?=?3)+
??scale_color_brewer(palette?=?"Set1")+
??theme_minimal()

rocs
plot of chunk unnamed-chunk-41

是不是很簡單呢？二分類資料常見的各種評(píng)價(jià)指標(biāo)都有了，圖也有了，還比較了多個(gè)模型，一舉多得，tidymodels，你值得擁有！

本文首發(fā)于公眾號(hào)：醫(yī)學(xué)和生信筆記

“

本文由 mdnice 多平臺(tái)發(fā)布

總結(jié)

以上是生活随笔為你收集整理的tidymodels搞定二分类资料多个模型评价和比较的全部內(nèi)容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網(wǎng)站內(nèi)容還不錯(cuò)，歡迎將生活随笔推薦給好友。

上一篇： c语言continue作用于,c语言co
下一篇： libcurl基础教程及封装（支持同步异