R/classification_models.R

In ahmedelmahy/RNASeqclassifier1: A collection of functions for feature selection and classifier development from RNASeq data.

#' fit a svmRadial
#'
#' @param x A train data.frame
#' @param y A vector of train classes
#' @param fitControl A configuration variable for caret
fit_svmRadial <- function(x, y, fitControl){
  fit_tune <- train(x = x,
                    y = y,
                    method = "svmRadial",
                    trControl = fitControl)

  fit <- train(x = x,
               y = y,
               method = "svmRadial",
               trControl = fitControl,
               tuneGrid=expand.grid(.sigma = fit_tune$bestTune$sigma,
                                    .C=fit_tune$bestTune$C),
               metric = "ROC")
  return(fit)
}

#' fit a svmLinear
#'
#' @param x A train data.frame
#' @param y A vector of train classes
#' @param fitControl A configuration variable for caret
fit_svmLinear <- function(x, y, fitControl){
  fit_tune <- train(x = x,
                    y = y,
                    method = "svmLinear",
                    trControl = fitControl)

  fit <- train(x = x,
               y = y,
               method = "svmLinear",
               trControl = fitControl,
               tuneGrid=expand.grid(.C=fit_tune$bestTune$C),
               metric = "ROC")
  return(fit)
}

#' fit a glmnet model
#'
#' @param x A train data.frame
#' @param y A vector of train classes
#' @param fitControl A configuration variable for caret
fit_glm <- function(x,y, fitControl){
  fit_tune <- train(x = x,
                    y = y,
                    method = "glmnet",
                    trControl = fitControl)

  fit <- caret::train(x = x,
                      y = y,
                      method ="glmnet",
                      tuneGrid=expand.grid(.alpha = fit_tune$bestTune$alpha,
                                           .lambda = fit_tune$bestTune$lambda),
                      trControl = fitControl)
  return(fit)
}

#' fit a random forest model
#'
#' @param x A train data.frame
#' @param y A vector of train classes
#' @param fitControl A configuration variable for caret
fit_rf <- function(x,y, fitControl){
  fit_tune <- train(x = x,
                    y = y,
                    method = "rf",
                    trControl = fitControl)

  fit <- caret::train(x = x,
                      y = y,
                      method ="rf",
                      tuneGrid=expand.grid(.mtry=fit_tune$bestTune$mtry),
                      trControl = fitControl)
  return(fit)
}

#' Create confusion matrix of the predictions vs reference class in the
#' test set.
#'
#' @param fit a caret model
#' @param test_df A test dataframe
#' @param test_class A vector of the classes
#' @return A confusion matrix
class_predict <- function(fit, test_df, test_class){
  predictions<-predict(object=fit,test_df)
  t <- confusionMatrix(predictions, test_class)
  return(t)
}