KNN chunk NBA

```{r knnModel}
library(pacman)
p_load(caret, ggplot)

knnTrain = train_final

knnTest = test_final

# Categoricals to factors
for (i in c(8:10)){
  knnTrain[,i] <- as.factor(knnTrain[,i])
  knnTest[,i] <- as.factor(knnTest[,i])
}

# Sample for faster performance
set.seed(7)
sample1 = sample(1:84616, 1000)

# Adding names to factors due to error when running train() "one of the class levels is not a valid R variable name"
levels(knnTrain$LOCATION_HOME) = c("A", "H")
levels(knnTrain$FGM) = c("miss", "made")
levels(knnTrain$PTS_TYPE_2) = c("three", "two")

levels(knnTest$LOCATION_HOME) = c("A", "H")
levels(knnTest$FGM) = c("miss", "made")
levels(knnTest$PTS_TYPE_2) = c("three", "two")

# Setting parameters for tuning repetitions and cross-validation in model
ctrl <- trainControl(
  classProbs = TRUE,
  method = "cv", # cross-validation method
  number = 5,    # number of folds
  )

# Modeling with training set
fgmknnModel <- train(FGM~., data = knnTrain, method = "knn", tuneGrid=expand.grid(k = c(200, 300)), trControl = ctrl)
fgmknnModel
summary(fgmknnModel)
# Determining and plotting the variable importance
fgmimpvar = varImp(fgmknnModel, scale = T)
ggplot(fgmimpvar)

# Predicting shot outcome and establishing final accuracy using test set
TestPred <- predict(fgmknnModel, knnTest)

# Creating the confusion matrix using the predicted and test values for FGM
confusionMatrix(TestPred, knnTest$FGM, positive = "made")

```