Update 2-modeling.R

ML/2-modeling.R

@@ -53,7 +53,6 @@ table(ds_test$ft4_dia) %>% prop.table()
 
 # random forest classification -----------------------------------------------------------
 
-# base model - No Hyper Tuning
 
 rf_recipe <- r$recipe(ft4_dia ~ . , data = ds_train) %>%
   r$step_rm(FT4) %>%
@@ -280,8 +279,8 @@ ds_reg_class_pred <- reg_test_results %>%
       ,TSH > 4.2 & `.pred` > 0.93 ~ "Non-Hypo"
       ,TSH < 0.27 & `.pred` > 1.7  ~ "Hyper"
       ,TSH < 0.27 & `.pred` < 1.7  ~ "Non-Hyper"
-    )
-  ) %>%
+      )
+    ) %>%
   dplyr::mutate(dplyr::across(
     ft4_dia_pred
     , ~factor(., levels = c("Hypo", "Non-Hypo","Hyper", "Non-Hyper")
@@ -302,8 +301,48 @@ ggplot(reg_test_results %>% tune::collect_predictions() , aes(x = FT4, y = .pred
 
 # check orginal data
 
-model_data %>% dplyr::group_by(ft4_dia) %>%
+model_data %>%
+  dplyr::mutate(tsh_level = ifelse(TSH > 4.2, "high", "low")) %>%
+  dplyr::group_by(tsh_level, ft4_dia) %>%
   dplyr::summarise(
     n = n()
+  ) %>%
+  mutate(freq = n / sum(n))
+
+
+
+# nnet reg ----------------------------------------------------------------
+
+normalized_rec <- recipes::recipe(FT4 ~ ., data = ds_train) %>%
+  recipes::step_rm(ft4_dia) %>%
+  recipes::step_impute_bag(recipes::all_predictors()) %>%
+  # recipes::step_corr(recipes::all_numeric_predictors()) %>%
+  recipes::step_normalize(recipes::all_numeric_predictors() , -anchor_age) %>%
+  recipes::step_dummy(gender)
+
+nnet_spec <-
+  p$mlp(hidden_units = tune(), penalty = tune(), epochs = tune()) %>%
+  p$set_engine("nnet", MaxNWts = 2600) %>%
+  p$set_mode("regression")
+
+nnet_param <-
+  nnet_spec %>%
+  tune$extract_parameter_set_dials() %>%
+  update(hidden_units = d$hidden_units(c(1, 27)))
+
+nnet_reg_workflow <- wf$workflow() %>%
+  wf$add_model(nnet_spec) %>%
+  wf$add_recipe(normalized_rec)
+
+data_fold_reg <- rsamp$vfold_cv(ds_train, v = 5)
+
+nnet_reg_tune <- nnet_reg_workflow %>%
+    tune::tune_grid(
+    data_fold_reg
+    ,grid = nnet_param %>% d$grid_regular()
+    ,verbose = TRUE
   )
 
+
+
+