---
title: "Stat. 652 - NHANES kNN"
author: "Prof. Eric A. Suess"
format: 
  html:
    embed-resources: true
---


Here we try using Tidymodels and Recipe to fit a classifier, the kNN model.

Note the need to impute values because of the high percentage of missing values.

For the kNN we use only numeric features.

```{r}
library(pacman)

p_load(NHANES, tidyverse, tidymodels, skimr, naniar)
```

### Step 1: Access the data.

```{r}
data(NHANES)

NHANES
```

### Step 2: Clean the data.  Prepare the data and make a recipe for applying the steps needed to preprocess the data.

First drop all of the rows where the y-variable SleepTrouble is missing.

Second select all of the numeric variables because we are using kNN which computes distances and needs to have numeric variables.  We create a list of the numeric variables and then use the one_of() function to pick out the columns with these names.

```{r}
skim(NHANES)
```



```{r}
NHANES <- NHANES |> 
  mutate(SleepTrouble = fct_relevel(SleepTrouble, "Yes")) 

NHANES_SleepTrouble <- NHANES |> select(-ID, -SleepHrsNight) |>
  select( SleepTrouble, everything() ) |>
  drop_na(SleepTrouble)
NHANES_SleepTrouble

```

Summarize the y-variable.  This is the Null Model.

```{r}
NHANES_SleepTrouble |> group_by(SleepTrouble) |>
  summarize(n = n()) |>
  mutate(freq = n / sum(n))
```

Make the first split with 80% of the data being in the trainning data set.

```{r}
NHANES_SleepTrouble_split <- initial_split(NHANES_SleepTrouble, prop = 0.8)
NHANES_SleepTrouble_split
```

Training data.

```{r}
NHANES_SleepTrouble_split |>
  training() 
```

```{r}
NHANES_SleepTrouble_split |>
  training() |>
  vis_miss()
```

```{r}
NHANES_SleepTrouble_split |>
  testing() 
```

```{r}
NHANES_SleepTrouble_split |>
  testing() |>
  vis_miss()
```

Create the *recipe()* for applying the preprocessing using *prep()*.  Note the use of *step_nzv()*, which removes any columns that have very low variability, and the use of the *step_impute_median()* function, which fills in the cells that are missing with the median of the column.

```{r}
NHANES_SleepTrouble_recipe <- training(NHANES_SleepTrouble_split) |>
  recipe(SleepTrouble ~ .) |>
  step_select(SleepTrouble, all_numeric()) |> 
  step_nzv(all_predictors()) |>
  step_impute_median(all_numeric()) |>
  step_center(all_numeric(), -all_outcomes()) |>
  step_scale(all_numeric(), -all_outcomes()) |>
  prep()

summary(NHANES_SleepTrouble_recipe)

tidy(NHANES_SleepTrouble_recipe)
```

Note that the variable SleepTrouble is moved to the last column of the dataset.

We "bake" the receipe into the Training data to see what the preprocesing does to the data.  We do the same for the Testing data.

```{r}
NHANES_SleepTrouble_training <- NHANES_SleepTrouble_recipe |>
  bake(training(NHANES_SleepTrouble_split)) 

NHANES_SleepTrouble_training
```


```{r}
NHANES_SleepTrouble_testing <- NHANES_SleepTrouble_recipe |>
  bake(testing(NHANES_SleepTrouble_split)) 

NHANES_SleepTrouble_testing
```


### Step 3: Training a model on the data

Setup the models.

```{r}
NHANES_SleepTrouble_kknn <- nearest_neighbor(neighbors = 21) |>
  set_engine("kknn") |>
  set_mode("classification") |>
  fit(SleepTrouble ~ ., data = NHANES_SleepTrouble_training)

NHANES_SleepTrouble_ranger <- rand_forest(trees = 100) |>
  set_engine("ranger") |>
  set_mode("classification") |>
  fit(SleepTrouble ~ ., data = NHANES_SleepTrouble_training)

NHANES_SleepTrouble_rf <- rand_forest(trees = 100) |>
  set_engine("randomForest") |>
  set_mode("classification") |>
  fit(SleepTrouble ~ ., data = NHANES_SleepTrouble_training)
```


```{r}
predict(NHANES_SleepTrouble_kknn, NHANES_SleepTrouble_testing)

predict(NHANES_SleepTrouble_ranger, NHANES_SleepTrouble_testing)

predict(NHANES_SleepTrouble_rf, NHANES_SleepTrouble_testing)
```

```{r}
NHANES_SleepTrouble_kknn |>
  predict(NHANES_SleepTrouble_testing) |>
  bind_cols(NHANES_SleepTrouble_testing) 

NHANES_SleepTrouble_ranger |>
  predict(NHANES_SleepTrouble_testing) |>
  bind_cols(NHANES_SleepTrouble_testing) 

NHANES_SleepTrouble_rf |>
  predict(NHANES_SleepTrouble_testing) |>
  bind_cols(NHANES_SleepTrouble_testing) 
```


### Step 4: Evaluate the models.

For the testing data.


```{r}
NHANES_SleepTrouble_kknn |>
  predict(NHANES_SleepTrouble_testing) |>
  bind_cols(NHANES_SleepTrouble_testing) |>
  metrics(truth = SleepTrouble, estimate = .pred_class)

NHANES_SleepTrouble_ranger |>
  predict(NHANES_SleepTrouble_testing) |>
  bind_cols(NHANES_SleepTrouble_testing) |>
  metrics(truth = SleepTrouble, estimate = .pred_class)

NHANES_SleepTrouble_rf |>
  predict(NHANES_SleepTrouble_testing) |>
  bind_cols(NHANES_SleepTrouble_testing) |>
  metrics(truth = SleepTrouble, estimate = .pred_class)
```

```{r}
NHANES_SleepTrouble_kknn |>
  predict(NHANES_SleepTrouble_testing) |>
  bind_cols(NHANES_SleepTrouble_testing) |>
  conf_mat(truth = SleepTrouble, estimate = .pred_class)

NHANES_SleepTrouble_ranger |>
  predict(NHANES_SleepTrouble_testing) |>
  bind_cols(NHANES_SleepTrouble_testing) |>
  conf_mat(truth = SleepTrouble, estimate = .pred_class)

NHANES_SleepTrouble_rf |>
  predict(NHANES_SleepTrouble_testing) |>
  bind_cols(NHANES_SleepTrouble_testing) |>
  conf_mat(truth = SleepTrouble, estimate = .pred_class)
```