diff --git a/Machine_Learning/New Tree Based pipelines/Extratrees_classification_PCA_autosplit.R b/Machine_Learning/New Tree Based pipelines/Extratrees_classification_PCA_autosplit.R
deleted file mode 100644
index d8aafd0..0000000
--- a/Machine_Learning/New Tree Based pipelines/Extratrees_classification_PCA_autosplit.R	
+++ /dev/null
@@ -1,196 +0,0 @@
-# First time running this pipeline you will need to install some R packages
-# 
-# Various R packages that are required
-library(readr)
-library(dplyr)
-library(stringr)
-library(ranger)
-library(lubridate)
-library(pROC)
-library(ggfortify)
-library(janitor)
-library(ggplot2)
-library(gridExtra)
-
-# Only the first iteration start here
-# Load in data that will be used to train
-#Autosplit data
-if (isFALSE(exists("random_forest_input"))){
-  setwd("C:\\Users\\file\\path")
-  data <- read.csv("clean_data.csv", check.names=FALSE)
-  data <- data[,-1]
-  data_size=nrow(data)
-  indexes=sample(1:nrow(data),size = 0.8*data_size)
-  training=data[indexes,]
-  validation=data[-indexes,]
-  
-  #This line was added to have the pipeline work with categorical data
-  training$Diagnosis = as.factor(training$Diagnosis)
-  # Load in validation dataset
-  validation$Diagnosis = as.factor(validation$Diagnosis)
-}
-
-#####################################################################
-##### All iterations after the first start here
-#####################################################################
-
-# Model building and feature selection will end when 3 columns aka 2 features are left
-while (as.numeric(ncol(training)) >3 ){
-  if (exists("random_forest_input")){
-    training <- random_forest_input
-  }
-  # Selecting tuning parameter based on the size of the data
-  # if() statement was added to account for iterations with very few ratios
-  training <- as.data.frame(training)
-  nvars <- floor(3*sqrt(dim(training)[2]))
-  if(nvars >= dim(training)[2]-1){nvars <- floor(sqrt(dim(training)[2]))}
-  
-  
-  # Create the model using ranger
-  mod <- ranger(data=training,           # The full dataset containing predictors and response variables
-                num.trees=min(30000,     # number of trees in the forest
-                length(training)),
-                splitrule = "extratrees", #If we use splitrule = "extratrees", replace = FALSE, sample.fraction = 1)  we are using Geurts et al. implementation.
-                replace = FALSE,
-                sample.fraction = 1,
-                max.depth = 8,
-                min.node.size = 3,
-                mtry=nvars,              # number of variables to sample for each split
-                importance='impurity',   # Type of variable importance measure to calculate
-                seed= 87,                # setting a seed for reproducibility
-                dependent.variable.name="Diagnosis",  
-                classification = TRUE
-                )
-  
-  # Now that we built our model lets check it's accuracy and save it
-  # Check accuracy
-  train_predictions <- predict(mod,training)
-  auc_train<- roc(as.numeric(training$Diagnosis), as.numeric(as.factor(train_predictions$predictions)))
-  
-  # Use below for more than two classes
-  #auc_train<- roc(as.numeric(training$Diagnosis), as.numeric(as.factor(mod$predictions)))
-  # Use if more than one category
-  #auc_train<- multiclass.roc(as.numeric(training$Stage), as.numeric(as.factor(mod$predictions)))
-  # View results
-  auc_train$auc
-  mod$confusion.matrix
-  
-  # Create prediction for validation dataset
-  # Calculate accuracy of validation predictions using auc
-  predictions <- predict(mod,validation)
-  auc_test<- roc(as.numeric(validation$Diagnosis), as.numeric(as.factor(predictions$predictions)))
-  auc_test$auc
-  
-  # For multiple classes
-  # test_auc_multi<- multiclass.roc(as.numeric(validation$Diagnosis), as.numeric(as.factor(predictions$predictions)))
-  # test_auc_multi$auc
-  
-  
-  # Add everything to list
-  if (isFALSE(exists("list_test_auc"))){
-    list_test_auc <- rbind(c( ncol(training), auc_test$auc))
-  }
-  if (exists("list_test_auc")){
-    list_test_auc <- rbind(list_test_auc, c(ncol(training), auc_test$auc))
-  }
-  # Create training auc list
-  if (isFALSE(exists("list_train_auc"))){
-    list_train_auc <- rbind(c( ncol(training), auc_train$auc))
-  }
-  if (exists("list_train_auc")){
-    list_train_auc <- rbind(list_train_auc, c(ncol(training), auc_train$auc))
-  }
-  
-  # Add predicton error to list
-  if (isFALSE(exists("list_prediction_error"))){
-    list_prediction_error <- rbind(c( ncol(training), mod$prediction.error))
-  }
-  if (exists("list_prediction_error")){
-    list_prediction_error <- rbind(list_prediction_error, c(ncol(training), mod$prediction.error))
-  }
-  
-  # Save feature importance list
-  csv_file_name <- paste0("importance", "_", as.numeric(ncol(training)), ".csv")
-  write.csv(mod$variable.importance, csv_file_name)
-  
-  
-  # Make PCA plot
-  # Remove zero variance columns
-  input<- training
-  input <- input %>%
-    clean_names()
-  input$Diagnosis = as.numeric(as.factor(training$Diagnosis))
-  input <- input[ , which(apply(input, 2, var) != 0)]
-  # Calculate PCA
-  pca_res <- prcomp(input, scale. = TRUE)
-  par(mar=c(5,6,4,2))
-  title <- paste0("Training ",as.numeric(ncol(training)))
-  plot1 <- autoplot(pca_res,
-                    data = training,
-                    colour = 'Diagnosis',
-                    main = title)
-  plot1<- plot1 + theme(legend.position = "bottom")
-  #print(plot1)
-
-  # Make PCA plot
-  # Remove zero variance columns
-  input<- validation
-  input <- input %>%
-    clean_names()
-  input$Diagnosis = as.numeric(as.factor(validation$Diagnosis))
-  input <- input[ , which(apply(input, 2, var) != 0)]
-  # Calculate PCA
-  pca_res <- prcomp(input, scale. = TRUE)
-  par(mfrow=c(1,1))
-  par(mar=c(5,6,4,2))
-  title <- paste0("Test ",as.numeric(ncol(training)))
-  plot2 <- autoplot(pca_res,
-                    data = validation,
-                    colour = 'Diagnosis',
-                    main = title)
-  plot2<- plot2 + theme(legend.position = "bottom")
-
-# Create the PCA plots
-  par(mfrow=c(1,2))
-  par(mar=c(5,6,4,2))
-  grid.arrange(plot1, plot2, nrow = 1)
-
-  
-  
-  # Recursive feature elimination
-  # Take the top 90% of features and create a new list
-  var_importance <- as.data.frame(mod$variable.importance)
-  ordered <- var_importance[order(-var_importance$`mod$variable.importance`), , drop = FALSE]
-  N <- nrow(ordered)
-  n <- ceiling(N*.9)
-  top90per <- ordered[1:n, ,drop = FALSE]
-  names <- unlist(row.names(top90per))
-  
-  # Take off the one worst feature if less than 10
-  if (nrow(ordered) < 10) {
-    ordered <- head(ordered, -1)
-    names <- unlist(row.names(ordered))
-  }
-  
-  random_forest_input <-training %>% 
-    select(Diagnosis,names)
-  
-  validation <-validation %>% 
-    select(Diagnosis,names)
-  
-  
-  # Save model
-  rds_file_name <- paste0("Diagnosis", as.numeric(ncol(training)), ".rds")
-  saveRDS(mod, file = rds_file_name)
-  
-  
-}
-
-##################################################
-##################################################
-# End of pipeline everything below is optional
-# Write summary of AUC
-write.csv(list_train_auc, "list_train_auc.csv") 
-write.csv(list_test_auc, "list_test_auc.csv") 
-
-