updates

ML/2-modeling.R

@@ -144,3 +144,25 @@ rf_reg_predict <- reg_train %>%
   dplyr::bind_cols(
     predict(rf_base_reg_fit, reg_train)
   )
+
+reg_metrics(rf_reg_predict, truth = FT4, estimate = .pred)
+
+rf_reg_tune_model <- p$rand_forest(trees = tune(), mtry = tune(), min_n = tune()) %>%
+  p$set_engine("ranger") %>% p$set_mode("regression")
+
+rf_reg_pred <- dplyr::select(reg_train, -FT4, -subject_id, -charttime)
+
+rf_reg_param <- p$extract_parameter_set_dials(rf_reg_tune_model) %>%
+  update(mtry = d$finalize(d$mtry(), rf_reg_pred))z
+
+data_fold_reg <- rsamp$vfold_cv(reg_train, v = 5)
+
+# takes around 1 hr to run grid search.  saving best params manaually
+rf_reg_tune <- rf_reg_workflow %>%
+    tune::tune_grid(
+    data_fold
+    ,grid = rf_reg_param %>% d$grid_regular()
+  )
+
+
+

chapter3.qmd

@@ -68,7 +68,8 @@ All data handling and modeling were performed using R and R Studio. The current
 #| label: tbl-platform-info
 #| tbl-cap: Session Info R Environment
 #| echo: false
-#| #| message: false
+#| message: false
+#| #| warning: false
 
 df_session_platform <-  devtools::session_info()$platform %>% 
   unlist(.) %>% 
@@ -90,6 +91,7 @@ knitr::kable(
 #| tbl-cap: Package Info R Environment
 #| echo: false
 #| message: false
+#| warning: false
 
 
 df_session_packages <-  devtools::session_info(include_base = TRUE)$packages %>% 
@@ -99,7 +101,7 @@ df_session_packages <-  devtools::session_info(include_base = TRUE)$packages %>%
 
 colnames(df_session_packages) <- c("Package", "Loaded Version", "Date")
 
-used_packages <-renv::dependencies() %>% dplyr::select(2)
+used_packages <-renv::dependencies(progress = FALSE) %>% dplyr::select(2)
 
 df_session_packages <- df_session_packages %>% 
   dplyr::filter(Package %in% used_packages$Package)