R/CustomFXs_Caret.R
In eisascience/scRNASeqEisa:
Defines functions
GetTraingStats
GetEnsmbleVarImportance
ClassifyCellsCustom
GetTraingStats <- function(ClassifierSet=NA, AvailableTrainConfMatrix="", AvailableTestConfMatrix=""){
tempDF1 <- data.frame(lapply(AvailableTrainConfMatrix, function(conftabN){
c(ClassifierSet[[conftabN]]$byClass[c(1:4, 6:10)],
ClassifierSet[[conftabN]]$overall[c(1,3,4)])
}))
colnames(tempDF1) <- AvailableTrainConfMatrix
tempDF2 <- data.frame(lapply(AvailableTestConfMatrix, function(conftabN){
c(ClassifierSet[[conftabN]]$byClass[c(1:4, 6:10)],
ClassifierSet[[conftabN]]$overall[c(1,3,4)])
}))
colnames(tempDF2) <- AvailableTestConfMatrix
data.frame(cbind(tempDF1, tempDF2))[,sort(c(AvailableTestConfMatrix, AvailableTrainConfMatrix))]
}
GetEnsmbleVarImportance <- function(ClassiferSet, AvailableClassifiers="", ScaleVarImp=T){
if(AvailableClassifiers[1]=="") AvailableClassifiers = names(ClassifierSet)
FoundVarImps <- lapply(AvailableClassifiers, function(clN){
print(clN)
tempOut <- try(varImp(ClassiferSet[[clN]],scale = ScaleVarImp), silent = T)
if(class(tempOut)[1]=="try-error") {
return(NULL)
print("No VarImp...")
} else{
tempOut2 <- tempOut$importance
if(ncol(tempOut2)<2){
tempOut2 <- data.frame(cbind(tempOut2, rep(NA, length(tempOut2))))
}
return(data.frame(t(tempOut2)))
}
})
names(FoundVarImps) <- AvailableClassifiers
FoundVarImps[sapply(FoundVarImps, is.null)] <- NULL
FoundVarImpsDF <- as.data.frame(data.table::rbindlist(FoundVarImps))
FoundVarImpsDF <- FoundVarImpsDF[is.odd(1:nrow(FoundVarImpsDF)),]
rownames(FoundVarImpsDF) <- names(FoundVarImps)
FoundVarImpsDF
}
ClassifyCellsCustom <- function(Classifier.rds.path = "", ClassifierNames="", testing.data, log10T=T, returnTraining=F){
if(file.exists(Classifier.rds.path)){
MultiClassifierResults <- readRDS(Classifier.rds.path)
Available.Classifiers <- names(MultiClassifierResults)[grepl("classifier", names(MultiClassifierResults))]
print(Available.Classifiers)
if(ClassifierNames[1]=="") ClassifierNames <- Available.Classifiers
if(length(which(ClassifierNames %in% Available.Classifiers)) < length(ClassifierNames)) {
print("Names in ClassifierNames did not match whats available")
ClassifierNames <- Available.Classifiers
}
MultiClassifier <- MultiClassifierResults[Available.Classifiers]
if(class(testing.data)[1]=="seurat") {
print("Converting Seurat Sparse Matrix to one with 0s .... ")
testing.data <- Matrix::as.matrix(t(testing.data@data))
} else {
if(!(class(testing.data)[1] %in% c("dgCMatrix", "matrix", "Matrix")) ) {
print("DGE mat not Seurat or dgCMatrix")
print("Make sure columns are cells and rows are genes and convert to expected format for speed")
testing.data <- Matrix::as.matrix((testing.data))
}
}
if(log10T) {
print("log-transforming")
testing.data <- log10(testing.data+1)
}
print("Starting Classification")
testing.data.yhat <- data.frame(lapply(Available.Classifiers, function(ClassifX){
print(ClassifX)
tempClassifier <- MultiClassifier[[ClassifX]]
y.hat <- predict(tempClassifier, newdata = testing.data)
return(y.hat)
}))
colnames(testing.data.yhat) <- Available.Classifiers
NotLevelName <- levels(testing.data.yhat[,1])[grep("Not", levels(testing.data.yhat[,1]))]
testing.data.yhat$CountNot <- apply(testing.data.yhat, 1, function(x) sum(grepl(NotLevelName, x)))
testing.data.yhat$CountTot <- rep(length(Available.Classifiers), nrow(testing.data.yhat))
testing.data.yhat$NotProb <- testing.data.yhat$CountNot/testing.data.yhat$CountTot
rownames(testing.data.yhat) <- rownames(testing.data)
return(list(yhat.DF = testing.data.yhat,
classification.levels = levels(testing.data.yhat[,1]),
Available.Classifiers = Available.Classifiers,
log10T = log10T,
DGEcolNames = colnames(testing.data),
DGErowNames = rownames(testing.data),
MultiClassifierResults = ifelse(returnTraining, MultiClassifierResults, "NULL")))
} else {
print("Check classifier path ....")
}
}
MultiClassifier_Cells_SERIII <- function (object_train,
object_test,
training.genes = NULL,
Y_Train_True = NULL,
Y_Test_True = NULL,
log10p1=T,
do_Adaboost = F,
do_Caret_LinSVM = F,
do_ranger_RF = F,
do_Caret_NaiveBayes = F,
do_Caret_NNet = F,
do_Caret_radialSVM = F,
do_ElasticNet = F,
do_tree = F,
do_tree_gini = F,
do_Caret_pcaNNet = F,
do_Caret_rangerRF = F,
do_glmStepAIC=F,
do_stepLDA=F,
crossValReps = 3,
NcrossVal = 10)
{
if(do_Adaboost) library(fastAdaboost)
if(do_Caret_LinSVM) library(caret)
if(do_Caret_NaiveBayes) library(caret)
if(do_Caret_NNet) library(caret)
if(do_ranger_RF) library(ranger)
if(do_ElasticNet) library(caret)
require(Matrix)
results_ls <- list()
training.classes <- as.vector(x = Y_Train_True)
training.genes <- SetIfNull(x = training.genes, default = rownames(x = object_train@assays$RNA@data))
training.data <- as.data.frame(x = as.matrix(x = Matrix::t(object_train@assays$RNA@data)))[,training.genes]
testing.classes <- as.vector(x = Y_Test_True)
testing.genes <- SetIfNull(x = training.genes, default = rownames(x = object_test@assays$RNA@data))
testing.data <- as.data.frame(x = as.matrix(x = Matrix::t(x = object_test@assays$RNA@data)))[,testing.genes ]
if(log10p1) training.data <- log10(training.data + 1)
if(log10p1) testing.data <- log10(testing.data + 1)
training.data$Class <- factor(x = training.classes)
#colnames(training.data) <- gsub("-", "", colnames(training.data))
#colnames(testing.data) <- gsub("-", "", colnames(testing.data))
results_ls$test_data <- testing.data
results_ls$test_Y <- Y_Test_True
results_ls$train_data <- training.data
results_ls$train_T <- Y_Train_True
#############template
# if(do_newModel){
#
#
# print("running newModel")
# TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
#
#
# newModel_classifier <- train(Class ~ ., data = training.data,
# method = "newModel",
# trControl= TrainingParameters,
# preProcess = c("scale","center"),
# na.action = na.omit)
#
#
# newModel_yhat <- predict(newModel_classifier, testing.data)
#
#
# results_ls$Caret_newModel_train_yhat <- predict(newModel_classifier, newdata=training.data[,which(colnames(training.data) !="Class")])
#
#
# results_ls$Caret_newModel_classifier <- newModel_classifier
# results_ls$Caret_newModel_test_yhat <- newModel_yhat
# results_ls$Caret_newModel_test_ConfMat <- confusionMatrix(newModel_yhat, (Y_Test_True))
# results_ls$Caret_newModel_test_ConfTab <- table(pred=newModel_yhat, truth=Y_Test_True)
#
# results_ls$Caret_newModel_train_ConfMat <- confusionMatrix((results_ls$Caret_newModel_train_yhat), (training.data$Class))
# results_ls$Caret_newModel_train_ConfTab <- table(pred=results_ls$Caret_newModel_train_yhat, truth=training.data$Class)
#
#
# colnames(training.data) <- colnames(results_ls$train_data)
# colnames(testing.data) <- colnames(results_ls$test_data)
#
#
# }
if(do_stepLDA){
print("running stepLDA")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
stepLDA_classifier <- train(Class ~ ., data = training.data,
method = "stepLDA",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit)
stepLDA_yhat <- predict(stepLDA_classifier, testing.data)
results_ls$Caret_stepLDA_train_yhat <- predict(stepLDA_classifier, newdata=training.data[,which(colnames(training.data) !="Class")])
results_ls$Caret_stepLDA_classifier <- stepLDA_classifier
results_ls$Caret_stepLDA_test_yhat <- stepLDA_yhat
results_ls$Caret_stepLDA_test_ConfMat <- confusionMatrix(stepLDA_yhat, (Y_Test_True))
results_ls$Caret_stepLDA_test_ConfTab <- table(pred=stepLDA_yhat, truth=Y_Test_True)
results_ls$Caret_stepLDA_train_ConfMat <- confusionMatrix((results_ls$Caret_stepLDA_train_yhat), (training.data$Class))
results_ls$Caret_stepLDA_train_ConfTab <- table(pred=results_ls$Caret_stepLDA_train_yhat, truth=training.data$Class)
colnames(training.data) <- colnames(results_ls$train_data)
colnames(testing.data) <- colnames(results_ls$test_data)
}
if(do_glmStepAIC){
print("running glmStepAIC")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
glmStepAIC_classifier <- train(Class ~ ., data = training.data,
method = "glmStepAIC",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit)
glmStepAIC_yhat <- predict(glmStepAIC_classifier, testing.data)
results_ls$Caret_glmStepAIC_train_yhat <- predict(glmStepAIC_classifier, newdata=training.data[,which(colnames(training.data) !="Class")])
results_ls$Caret_glmStepAIC_classifier <- glmStepAIC_classifier
results_ls$Caret_glmStepAIC_test_yhat <- glmStepAIC_yhat
results_ls$Caret_glmStepAIC_test_ConfMat <- confusionMatrix(glmStepAIC_yhat, (Y_Test_True))
results_ls$Caret_glmStepAIC_test_ConfTab <- table(pred=glmStepAIC_yhat, truth=Y_Test_True)
results_ls$Caret_glmStepAIC_train_ConfMat <- confusionMatrix((results_ls$Caret_glmStepAIC_train_yhat), (training.data$Class))
results_ls$Caret_glmStepAIC_train_ConfTab <- table(pred=results_ls$Caret_glmStepAIC_train_yhat, truth=training.data$Class)
colnames(training.data) <- colnames(results_ls$train_data)
colnames(testing.data) <- colnames(results_ls$test_data)
}
if(do_tree){
library(caret)
library(rpart)
library(rpart.plot)
print("running tree based class with rpart")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
rpart_classifier <- train(x=training.data[,setdiff(colnames(training.data), c("Class"))],
y=training.data$Class,
method = "rpart",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit,
parms = list(split = "information"),
tuneLength = 10)
prp(rpart_classifier$finalModel, box.palette = "Blues", tweak = 1.2)
results_ls$Caret_rpart_treeViz <- recordPlot()
rpart_yhat <- predict(rpart_classifier, testing.data)
results_ls$Caret_rpart_train_yhat <- predict(rpart_classifier, newdata=training.data[,which(colnames(training.data) !="Class")])
results_ls$Caret_rpart_classifier <- rpart_classifier
results_ls$Caret_rpart_test_yhat <- rpart_yhat
results_ls$Caret_rpart_test_ConfMat <- confusionMatrix(rpart_yhat, (Y_Test_True))
results_ls$Caret_rpart_test_ConfTab <- table(pred=rpart_yhat, truth=Y_Test_True)
results_ls$Caret_rpart_train_ConfMat <- confusionMatrix((results_ls$Caret_rpart_train_yhat), (training.data$Class))
results_ls$Caret_rpart_train_ConfTab <- table(pred=results_ls$Caret_rpart_train_yhat, truth=training.data$Class)
colnames(training.data) <- colnames(results_ls$train_data)
colnames(testing.data) <- colnames(results_ls$test_data)
}
if(do_tree_gini){
library(caret)
library(rpart)
library(rpart.plot)
print("running tree based class with rpart w/ gini index")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
rpartGini_classifier <- train(x=training.data[,setdiff(colnames(training.data), c("Class"))],
y=training.data$Class,
method = "rpart",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit,
parms = list(split = "gini"),
tuneLength = 10)
prp(rpartGini_classifier$finalModel, box.palette = "Blues", tweak = 1.2)
results_ls$Caret_rpartGini_treeViz <- recordPlot()
rpartGini_yhat <- predict(rpartGini_classifier, testing.data)
results_ls$Caret_rpartGini_train_yhat <- predict(rpartGini_classifier, newdata=training.data[,which(colnames(training.data) !="Class")])
results_ls$Caret_rpartGini_classifier <- rpartGini_classifier
results_ls$Caret_rpartGini_test_yhat <- rpartGini_yhat
results_ls$Caret_rpartGini_test_ConfMat <- confusionMatrix(rpartGini_yhat, (Y_Test_True))
results_ls$Caret_rpartGini_test_ConfTab <- table(pred=rpartGini_yhat, truth=Y_Test_True)
results_ls$Caret_rpartGini_train_ConfMat <- confusionMatrix((results_ls$Caret_rpartGini_train_yhat), (training.data$Class))
results_ls$Caret_rpartGini_train_ConfTab <- table(pred=results_ls$Caret_rpartGini_train_yhat, truth=training.data$Class)
results_ls
colnames(training.data) <- colnames(results_ls$train_data)
colnames(testing.data) <- colnames(results_ls$test_data)
}
if(do_ElasticNet){
library(glmnet)
print("running ElasticNet Reg.")
#print(class(training.data))
#print(head(training.data))
#print((training.data$Class))
#print(levels(training.data$Class))
#print(colnames(training.data[,setdiff(colnames(training.data), c("Class"))]))
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps,
returnResamp="all",
classProbs=TRUE)
training.data.Y <- factor(as.character(training.data$Class))
training.data.X <- training.data[,setdiff(colnames(training.data), c("Class"))]
ElasticNet_classifier <- try(caret::train(training.data.X, training.data.Y,
method = "glmnet",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit), silent = T)
#print(class(ElasticNet_classifier))
if(class(ElasticNet_classifier)[1]=="try-error") {
print("first model failed, tring with alternate alpha and lambda...")
ElasticNet_classifier <- try(caret::train(training.data.X, training.data.Y,
method = "glmnet",
metric = "ROC",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit,
tuneGrid = expand.grid(alpha = c(0.1, 1),
lambda = c(0.001, 0.0001))),
silent = T)
}
if(class(ElasticNet_classifier)[1]=="try-error") {
warning("ElasticNet classifier failed ")
} else {
ElasticNet_yhat <- predict(ElasticNet_classifier, testing.data)
results_ls$Caret_ElasticNet_train_yhat <- predict(ElasticNet_classifier, newdata=training.data[,training.genes])
results_ls$Caret_ElasticNet_classifier <- ElasticNet_classifier
results_ls$Caret_ElasticNet_test_yhat <- ElasticNet_yhat
results_ls$Caret_ElasticNet_test_ConfMat <- confusionMatrix(ElasticNet_yhat, (Y_Test_True))
results_ls$Caret_ElasticNet_test_ConfTab <- table(pred=ElasticNet_yhat, truth=Y_Test_True)
results_ls$Caret_ElasticNet_train_ConfMat <- confusionMatrix((results_ls$Caret_ElasticNet_train_yhat), (training.data$Class))
results_ls$Caret_ElasticNet_train_ConfTab <- table(pred=results_ls$Caret_ElasticNet_train_yhat, truth=training.data$Class)
}
}
if(do_Adaboost){
print("running Adaboost")
Adaboost_classifier <- fastAdaboost::adaboost(Class ~ ., training.data, NcrossVal)
y_hat_Adaboost <- predict(Adaboost_classifier, newdata=testing.data)
results_ls$Adaboost_train_yhat <- predict(Adaboost_classifier, newdata=training.data[,training.genes])
results_ls$Adaboost_classifier <- Adaboost_classifier
results_ls$Adaboost_test_yhat <- y_hat_Adaboost
results_ls$Adaboost_test_ConfMat <- confusionMatrix((y_hat_Adaboost$class), (Y_Test_True))
results_ls$Adaboost_test_ConfTab <- table(pred=y_hat_Adaboost$class, truth=Y_Test_True)
results_ls$Adaboost_train_ConfMat <- confusionMatrix((results_ls$Adaboost_train_yhat$class), (training.data$Class))
results_ls$Adaboost_train_ConfTab <- table(pred=results_ls$Adaboost_train_yhat$class, truth=training.data$Class)
}
if(do_Caret_LinSVM) {
# training model with SVM
print("running Caret LinSVM")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
SVM_linear_classifier <- train(Class ~ ., data = training.data,
method = "svmLinear",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit)
SVM_Lin_yhat <- predict(SVM_linear_classifier, testing.data)
results_ls$Caret_LinSVM_train_yhat <- predict(SVM_linear_classifier, newdata=training.data[,training.genes])
results_ls$Caret_LinSVM_classifier <- SVM_linear_classifier
results_ls$Caret_LinSVM_test_yhat <- SVM_Lin_yhat
results_ls$Caret_LinSVM_test_ConfMat <- confusionMatrix(SVM_Lin_yhat, (Y_Test_True))
results_ls$Caret_LinSVM_test_ConfTab <- table(pred=SVM_Lin_yhat, truth=Y_Test_True)
results_ls$Caret_LinSVM_train_ConfMat <- confusionMatrix((results_ls$Caret_LinSVM_train_yhat), (training.data$Class))
results_ls$Caret_LinSVM_train_ConfTab <- table(pred=results_ls$Caret_LinSVM_train_yhat, truth=training.data$Class)
}
if(do_Caret_radialSVM) {
# training model with SVM
print("running Caret radialSVM")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
SVM_radial_classifier <- train(Class ~ ., data = training.data,
method = "svmRadial",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit)
SVM_Radial_yhat <- predict(SVM_radial_classifier, testing.data)
results_ls$Caret_radialSVM_train_yhat <- predict(SVM_radial_classifier, newdata=training.data[,training.genes])
results_ls$Caret_radialSVM_classifier <- SVM_radial_classifier
results_ls$Caret_radialSVM_test_yhat <- SVM_Radial_yhat
results_ls$Caret_radialSVM_test_ConfMat <- confusionMatrix(SVM_Radial_yhat, (Y_Test_True))
results_ls$Caret_radialSVM_test_ConfTab <- table(pred=SVM_Radial_yhat, truth=Y_Test_True)
results_ls$Caret_radialSVM_train_ConfMat <- confusionMatrix((results_ls$Caret_radialSVM_train_yhat), (training.data$Class))
results_ls$Caret_radialSVM_train_ConfTab <- table(pred=results_ls$Caret_radialSVM_train_yhat, truth=training.data$Class)
}
if(do_Caret_NaiveBayes) {
print("running Naive Bayes")
# training model with SVM
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
#Naive Bayes algorithm
NaiveBayes_classifier <- train(Class ~ ., data = training.data,
method = "nb",
preProcess=c("scale","center"),
trControl= TrainingParameters,
na.action = na.omit, silent = T)
NaiveBayes_yhat <- predict(NaiveBayes_classifier, testing.data)
results_ls$NaiveBayes_train_yhat <- predict(NaiveBayes_classifier, newdata=training.data[,training.genes])
results_ls$NaiveBayes_classifier <- NaiveBayes_classifier
results_ls$NaiveBayes_test_yhat <- NaiveBayes_yhat
results_ls$NaiveBayes_test_ConfMat <- confusionMatrix(NaiveBayes_yhat, (Y_Test_True))
results_ls$NaiveBayes_test_ConfTab <- table(pred=NaiveBayes_yhat, truth=Y_Test_True)
results_ls$NaiveBayes_train_ConfMat <- confusionMatrix((results_ls$NaiveBayes_train_yhat), (training.data$Class))
results_ls$NaiveBayes_train_ConfTab <- table(pred=results_ls$NaiveBayes_train_yhat, truth=training.data$Class)
}
if(do_Caret_NNet) {
print("running NNet")
# training model with SVM
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
#Naive Bayes algorithm
NNet_classifier <- train(Class ~ ., data = training.data,
method = "nnet",
preProcess=c("scale","center"),
trControl= TrainingParameters,
na.action = na.omit)
NNet_yhat <- predict(NNet_classifier, testing.data)
results_ls$NNet_train_yhat <- predict(NNet_classifier, newdata=training.data[,training.genes])
results_ls$NNet_classifier <- NNet_classifier
results_ls$NNet_test_yhat <- NNet_yhat
results_ls$NNet_test_ConfMat <- confusionMatrix(NNet_yhat, (Y_Test_True))
results_ls$NNet_test_ConfTab <- table(pred=NNet_yhat, truth=Y_Test_True)
results_ls$NNet_train_ConfMat <- confusionMatrix((results_ls$NNet_train_yhat), (training.data$Class))
results_ls$NNet_train_ConfTab <- table(pred=results_ls$NNet_train_yhat, truth=training.data$Class)
}
if(do_ranger_RF) {
# training
print("running ranger-RF")
RRF_classifier <- ranger::ranger(data = training.data,
dependent.variable.name = "Class",
classification = TRUE,
write.forest = TRUE)
RRF_yhat <- predict(RRF_classifier, testing.data)
results_ls$RRF_train_yhat <- predict(RRF_classifier, data=training.data[,training.genes])
results_ls$RRF_classifier <- RRF_classifier
results_ls$RRF_test_yhat <- RRF_yhat
results_ls$RRF_test_ConfMat <- confusionMatrix(RRF_yhat$predictions, (Y_Test_True))
results_ls$RRF_test_ConfTab <- table(pred=RRF_yhat$predictions, truth=Y_Test_True)
results_ls$RRF_train_ConfMat <- confusionMatrix((results_ls$RRF_train_yhat$predictions), (training.data$Class))
results_ls$RRF_train_ConfTab <- table(pred=results_ls$RRF_train_yhat$predictions, truth=training.data$Class)
}
if(do_Caret_rangerRF) {
# training model with SVM
print("running Caret RF::ranger ")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
rangerRFCaret_classifier <- train(Class ~ ., data = training.data,
method = "ranger",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit)
rangerRFCaret_yhat <- predict(rangerRFCaret_classifier, testing.data)
results_ls$Caret_RFranger_train_yhat <- predict(rangerRFCaret_classifier, newdata=training.data[,training.genes])
results_ls$Caret_RFranger_classifier <- rangerRFCaret_classifier
results_ls$Caret_RFranger_test_yhat <- rangerRFCaret_yhat
results_ls$Caret_RFranger_test_ConfMat <- confusionMatrix(rangerRFCaret_yhat, (Y_Test_True))
results_ls$Caret_RFranger_test_ConfTab <- table(pred=rangerRFCaret_yhat, truth=Y_Test_True)
results_ls$Caret_RFranger_train_ConfMat <- confusionMatrix((results_ls$Caret_RFranger_train_yhat), (training.data$Class))
results_ls$Caret_RFranger_train_ConfTab <- table(pred=results_ls$Caret_RFranger_train_yhat, truth=training.data$Class)
}
if(do_Caret_pcaNNet) {
# training model with SVM
print("running Caret pcaNNet ")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
pcaNNetCaret_classifier <- train(Class ~ ., data = training.data,
method = "pcaNNet",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit)
pcaNNetCaret_yhat <- predict(pcaNNetCaret_classifier, testing.data)
results_ls$Caret_pcaNNet_train_yhat <- predict(pcaNNetCaret_classifier, newdata=training.data[,training.genes])
results_ls$Caret_pcaNNet_classifier <- pcaNNetCaret_classifier
results_ls$Caret_pcaNNet_test_yhat <- pcaNNetCaret_yhat
results_ls$Caret_pcaNNet_test_ConfMat <- confusionMatrix(pcaNNetCaret_yhat, (Y_Test_True))
results_ls$Caret_pcaNNet_test_ConfTab <- table(pred=pcaNNetCaret_yhat, truth=Y_Test_True)
results_ls$Caret_RFranger_train_ConfMat <- confusionMatrix((results_ls$Caret_pcaNNet_train_yhat), (training.data$Class))
results_ls$Caret_RFranger_train_ConfTab <- table(pred=results_ls$Caret_pcaNNet_train_yhat, truth=training.data$Class)
}
classifierNames <- names(results_ls)[grep("classifier", names(results_ls))]
results_ls$ReSamp <- resamples(lapply(classifierNames, function(clsf){
results_ls[[clsf]]
}))
results_ls$ReSampSum <- summary(results_ls$ReSamp)
bwplot(results_ls$ReSamp)
results_ls$ReSampbwplot <- recordPlot()
dotplot(results_ls$ReSamp)
results_ls$ReSampdotplot <- recordPlot()
results_ls$modelCor <- modelCor(results_ls$ReSamp)
splom(results_ls$ReSamp)
results_ls$ReSampsplom <- recordPlot()
return(results_ls)
}
MultiClassifier_Cells <- function (object_train,
object_test,
training.genes = NULL,
Y_Train_True = NULL,
Y_Test_True = NULL,
log10p1=T,
do_Adaboost = F,
do_Caret_LinSVM = F,
do_ranger_RF = F,
do_Caret_NaiveBayes = F,
do_Caret_NNet = F,
do_Caret_radialSVM = F,
do_ElasticNet = F,
do_tree = F,
do_tree_gini = F,
do_Caret_pcaNNet = F,
do_Caret_rangerRF = F,
do_glmStepAIC=F,
do_stepLDA=F,
crossValReps = 3,
NcrossVal = 10)
{
if(do_Adaboost) library(fastAdaboost)
if(do_Caret_LinSVM) library(caret)
if(do_Caret_NaiveBayes) library(caret)
if(do_Caret_NNet) library(caret)
if(do_ranger_RF) library(ranger)
if(do_ElasticNet) library(caret)
require(Matrix)
results_ls <- list()
training.classes <- as.vector(x = Y_Train_True)
training.genes <- SetIfNull(x = training.genes, default = rownames(x = object_train@assays$RNA@data))
training.data <- as.data.frame(x = as.matrix(x = Matrix::t(object_train@assays$RNA@data)))[,training.genes]
testing.classes <- as.vector(x = Y_Test_True)
testing.genes <- SetIfNull(x = training.genes, default = rownames(x = object_test@assays$RNA@data))
testing.data <- as.data.frame(x = as.matrix(x = Matrix::t(x = object_test@assays$RNA@data)))[,testing.genes ]
if(log10p1) training.data <- log10(training.data + 1)
if(log10p1) testing.data <- log10(testing.data + 1)
training.data$Class <- factor(x = training.classes)
#colnames(training.data) <- gsub("-", "", colnames(training.data))
#colnames(testing.data) <- gsub("-", "", colnames(testing.data))
results_ls$test_data <- testing.data
results_ls$test_Y <- Y_Test_True
results_ls$train_data <- training.data
results_ls$train_T <- Y_Train_True
#############template
# if(do_newModel){
#
#
# print("running newModel")
# TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
#
#
# newModel_classifier <- train(Class ~ ., data = training.data,
# method = "newModel",
# trControl= TrainingParameters,
# preProcess = c("scale","center"),
# na.action = na.omit)
#
#
# newModel_yhat <- predict(newModel_classifier, testing.data)
#
#
# results_ls$Caret_newModel_train_yhat <- predict(newModel_classifier, newdata=training.data[,which(colnames(training.data) !="Class")])
#
#
# results_ls$Caret_newModel_classifier <- newModel_classifier
# results_ls$Caret_newModel_test_yhat <- newModel_yhat
# results_ls$Caret_newModel_test_ConfMat <- confusionMatrix(newModel_yhat, (Y_Test_True))
# results_ls$Caret_newModel_test_ConfTab <- table(pred=newModel_yhat, truth=Y_Test_True)
#
# results_ls$Caret_newModel_train_ConfMat <- confusionMatrix((results_ls$Caret_newModel_train_yhat), (training.data$Class))
# results_ls$Caret_newModel_train_ConfTab <- table(pred=results_ls$Caret_newModel_train_yhat, truth=training.data$Class)
#
#
# colnames(training.data) <- colnames(results_ls$train_data)
# colnames(testing.data) <- colnames(results_ls$test_data)
#
#
# }
if(do_stepLDA){
print("running stepLDA")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
stepLDA_classifier <- train(Class ~ ., data = training.data,
method = "stepLDA",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit)
stepLDA_yhat <- predict(stepLDA_classifier, testing.data)
results_ls$Caret_stepLDA_train_yhat <- predict(stepLDA_classifier, newdata=training.data[,which(colnames(training.data) !="Class")])
results_ls$Caret_stepLDA_classifier <- stepLDA_classifier
results_ls$Caret_stepLDA_test_yhat <- stepLDA_yhat
results_ls$Caret_stepLDA_test_ConfMat <- confusionMatrix(stepLDA_yhat, (Y_Test_True))
results_ls$Caret_stepLDA_test_ConfTab <- table(pred=stepLDA_yhat, truth=Y_Test_True)
results_ls$Caret_stepLDA_train_ConfMat <- confusionMatrix((results_ls$Caret_stepLDA_train_yhat), (training.data$Class))
results_ls$Caret_stepLDA_train_ConfTab <- table(pred=results_ls$Caret_stepLDA_train_yhat, truth=training.data$Class)
colnames(training.data) <- colnames(results_ls$train_data)
colnames(testing.data) <- colnames(results_ls$test_data)
}
if(do_glmStepAIC){
print("running glmStepAIC")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
glmStepAIC_classifier <- train(Class ~ ., data = training.data,
method = "glmStepAIC",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit)
glmStepAIC_yhat <- predict(glmStepAIC_classifier, testing.data)
results_ls$Caret_glmStepAIC_train_yhat <- predict(glmStepAIC_classifier, newdata=training.data[,which(colnames(training.data) !="Class")])
results_ls$Caret_glmStepAIC_classifier <- glmStepAIC_classifier
results_ls$Caret_glmStepAIC_test_yhat <- glmStepAIC_yhat
results_ls$Caret_glmStepAIC_test_ConfMat <- confusionMatrix(glmStepAIC_yhat, (Y_Test_True))
results_ls$Caret_glmStepAIC_test_ConfTab <- table(pred=glmStepAIC_yhat, truth=Y_Test_True)
results_ls$Caret_glmStepAIC_train_ConfMat <- confusionMatrix((results_ls$Caret_glmStepAIC_train_yhat), (training.data$Class))
results_ls$Caret_glmStepAIC_train_ConfTab <- table(pred=results_ls$Caret_glmStepAIC_train_yhat, truth=training.data$Class)
colnames(training.data) <- colnames(results_ls$train_data)
colnames(testing.data) <- colnames(results_ls$test_data)
}
if(do_tree){
library(caret)
library(rpart)
library(rpart.plot)
print("running tree based class with rpart")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
rpart_classifier <- train(x=training.data[,setdiff(colnames(training.data), c("Class"))],
y=training.data$Class,
method = "rpart",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit,
parms = list(split = "information"),
tuneLength = 10)
prp(rpart_classifier$finalModel, box.palette = "Blues", tweak = 1.2)
results_ls$Caret_rpart_treeViz <- recordPlot()
rpart_yhat <- predict(rpart_classifier, testing.data)
results_ls$Caret_rpart_train_yhat <- predict(rpart_classifier, newdata=training.data[,which(colnames(training.data) !="Class")])
results_ls$Caret_rpart_classifier <- rpart_classifier
results_ls$Caret_rpart_test_yhat <- rpart_yhat
results_ls$Caret_rpart_test_ConfMat <- confusionMatrix(rpart_yhat, (Y_Test_True))
results_ls$Caret_rpart_test_ConfTab <- table(pred=rpart_yhat, truth=Y_Test_True)
results_ls$Caret_rpart_train_ConfMat <- confusionMatrix((results_ls$Caret_rpart_train_yhat), (training.data$Class))
results_ls$Caret_rpart_train_ConfTab <- table(pred=results_ls$Caret_rpart_train_yhat, truth=training.data$Class)
colnames(training.data) <- colnames(results_ls$train_data)
colnames(testing.data) <- colnames(results_ls$test_data)
}
if(do_tree_gini){
library(caret)
library(rpart)
library(rpart.plot)
print("running tree based class with rpart w/ gini index")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
rpartGini_classifier <- train(x=training.data[,setdiff(colnames(training.data), c("Class"))],
y=training.data$Class,
method = "rpart",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit,
parms = list(split = "gini"),
tuneLength = 10)
prp(rpartGini_classifier$finalModel, box.palette = "Blues", tweak = 1.2)
results_ls$Caret_rpartGini_treeViz <- recordPlot()
rpartGini_yhat <- predict(rpartGini_classifier, testing.data)
results_ls$Caret_rpartGini_train_yhat <- predict(rpartGini_classifier, newdata=training.data[,which(colnames(training.data) !="Class")])
results_ls$Caret_rpartGini_classifier <- rpartGini_classifier
results_ls$Caret_rpartGini_test_yhat <- rpartGini_yhat
results_ls$Caret_rpartGini_test_ConfMat <- confusionMatrix(rpartGini_yhat, (Y_Test_True))
results_ls$Caret_rpartGini_test_ConfTab <- table(pred=rpartGini_yhat, truth=Y_Test_True)
results_ls$Caret_rpartGini_train_ConfMat <- confusionMatrix((results_ls$Caret_rpartGini_train_yhat), (training.data$Class))
results_ls$Caret_rpartGini_train_ConfTab <- table(pred=results_ls$Caret_rpartGini_train_yhat, truth=training.data$Class)
results_ls
colnames(training.data) <- colnames(results_ls$train_data)
colnames(testing.data) <- colnames(results_ls$test_data)
}
if(do_ElasticNet){
library(glmnet)
print("running ElasticNet Reg.")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
ElasticNet_classifier <- try(train(x=training.data[,setdiff(colnames(training.data), c("Class"))],
y=training.data$Class,
method = "glmnet",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit), silent = T)
if(class(ElasticNet_classifier)[1]=="try-error") {
ElasticNet_classifier <- try(train(x=training.data[,setdiff(colnames(training.data), c("Class"))],
y=training.data$Class,
method = "glmnet",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit,
tuneGrid = expand.grid(alpha = 0.1, lambda =0.01)),
silent = T)
}
ElasticNet_yhat <- predict(ElasticNet_classifier, testing.data)
results_ls$Caret_ElasticNet_train_yhat <- predict(ElasticNet_classifier, newdata=training.data[,training.genes])
results_ls$Caret_ElasticNet_classifier <- ElasticNet_classifier
results_ls$Caret_ElasticNet_test_yhat <- ElasticNet_yhat
results_ls$Caret_ElasticNet_test_ConfMat <- confusionMatrix(ElasticNet_yhat, (Y_Test_True))
results_ls$Caret_ElasticNet_test_ConfTab <- table(pred=ElasticNet_yhat, truth=Y_Test_True)
results_ls$Caret_ElasticNet_train_ConfMat <- confusionMatrix((results_ls$Caret_ElasticNet_train_yhat), (training.data$Class))
results_ls$Caret_ElasticNet_train_ConfTab <- table(pred=results_ls$Caret_ElasticNet_train_yhat, truth=training.data$Class)
}
if(do_Adaboost){
print("running Adaboost")
Adaboost_classifier <- fastAdaboost::adaboost(Class ~ ., training.data, NcrossVal)
y_hat_Adaboost <- predict(Adaboost_classifier, newdata=testing.data)
results_ls$Adaboost_train_yhat <- predict(Adaboost_classifier, newdata=training.data[,training.genes])
results_ls$Adaboost_classifier <- Adaboost_classifier
results_ls$Adaboost_test_yhat <- y_hat_Adaboost
results_ls$Adaboost_test_ConfMat <- confusionMatrix((y_hat_Adaboost$class), (Y_Test_True))
results_ls$Adaboost_test_ConfTab <- table(pred=y_hat_Adaboost$class, truth=Y_Test_True)
results_ls$Adaboost_train_ConfMat <- confusionMatrix((results_ls$Adaboost_train_yhat$class), (training.data$Class))
results_ls$Adaboost_train_ConfTab <- table(pred=results_ls$Adaboost_train_yhat$class, truth=training.data$Class)
}
if(do_Caret_LinSVM) {
# training model with SVM
print("running Caret LinSVM")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
SVM_linear_classifier <- train(Class ~ ., data = training.data,
method = "svmLinear",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit)
SVM_Lin_yhat <- predict(SVM_linear_classifier, testing.data)
results_ls$Caret_LinSVM_train_yhat <- predict(SVM_linear_classifier, newdata=training.data[,training.genes])
results_ls$Caret_LinSVM_classifier <- SVM_linear_classifier
results_ls$Caret_LinSVM_test_yhat <- SVM_Lin_yhat
results_ls$Caret_LinSVM_test_ConfMat <- confusionMatrix(SVM_Lin_yhat, (Y_Test_True))
results_ls$Caret_LinSVM_test_ConfTab <- table(pred=SVM_Lin_yhat, truth=Y_Test_True)
results_ls$Caret_LinSVM_train_ConfMat <- confusionMatrix((results_ls$Caret_LinSVM_train_yhat), (training.data$Class))
results_ls$Caret_LinSVM_train_ConfTab <- table(pred=results_ls$Caret_LinSVM_train_yhat, truth=training.data$Class)
}
if(do_Caret_radialSVM) {
# training model with SVM
print("running Caret radialSVM")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
SVM_radial_classifier <- train(Class ~ ., data = training.data,
method = "svmRadial",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit)
SVM_Radial_yhat <- predict(SVM_radial_classifier, testing.data)
results_ls$Caret_radialSVM_train_yhat <- predict(SVM_radial_classifier, newdata=training.data[,training.genes])
results_ls$Caret_radialSVM_classifier <- SVM_radial_classifier
results_ls$Caret_radialSVM_test_yhat <- SVM_Radial_yhat
results_ls$Caret_radialSVM_test_ConfMat <- confusionMatrix(SVM_Radial_yhat, (Y_Test_True))
results_ls$Caret_radialSVM_test_ConfTab <- table(pred=SVM_Radial_yhat, truth=Y_Test_True)
results_ls$Caret_radialSVM_train_ConfMat <- confusionMatrix((results_ls$Caret_radialSVM_train_yhat), (training.data$Class))
results_ls$Caret_radialSVM_train_ConfTab <- table(pred=results_ls$Caret_radialSVM_train_yhat, truth=training.data$Class)
}
if(do_Caret_NaiveBayes) {
print("running Naive Bayes")
# training model with SVM
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
#Naive Bayes algorithm
NaiveBayes_classifier <- train(Class ~ ., data = training.data,
method = "nb",
preProcess=c("scale","center"),
trControl= TrainingParameters,
na.action = na.omit, silent = T)
NaiveBayes_yhat <- predict(NaiveBayes_classifier, testing.data)
results_ls$NaiveBayes_train_yhat <- predict(NaiveBayes_classifier, newdata=training.data[,training.genes])
results_ls$NaiveBayes_classifier <- NaiveBayes_classifier
results_ls$NaiveBayes_test_yhat <- NaiveBayes_yhat
results_ls$NaiveBayes_test_ConfMat <- confusionMatrix(NaiveBayes_yhat, (Y_Test_True))
results_ls$NaiveBayes_test_ConfTab <- table(pred=NaiveBayes_yhat, truth=Y_Test_True)
results_ls$NaiveBayes_train_ConfMat <- confusionMatrix((results_ls$NaiveBayes_train_yhat), (training.data$Class))
results_ls$NaiveBayes_train_ConfTab <- table(pred=results_ls$NaiveBayes_train_yhat, truth=training.data$Class)
}
if(do_Caret_NNet) {
print("running NNet")
# training model with SVM
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
#Naive Bayes algorithm
NNet_classifier <- train(Class ~ ., data = training.data,
method = "nnet",
preProcess=c("scale","center"),
trControl= TrainingParameters,
na.action = na.omit)
NNet_yhat <- predict(NNet_classifier, testing.data)
results_ls$NNet_train_yhat <- predict(NNet_classifier, newdata=training.data[,training.genes])
results_ls$NNet_classifier <- NNet_classifier
results_ls$NNet_test_yhat <- NNet_yhat
results_ls$NNet_test_ConfMat <- confusionMatrix(NNet_yhat, (Y_Test_True))
results_ls$NNet_test_ConfTab <- table(pred=NNet_yhat, truth=Y_Test_True)
results_ls$NNet_train_ConfMat <- confusionMatrix((results_ls$NNet_train_yhat), (training.data$Class))
results_ls$NNet_train_ConfTab <- table(pred=results_ls$NNet_train_yhat, truth=training.data$Class)
}
if(do_ranger_RF) {
# training
print("running ranger-RF")
RRF_classifier <- ranger::ranger(data = training.data,
dependent.variable.name = "Class",
classification = TRUE,
write.forest = TRUE)
RRF_yhat <- predict(RRF_classifier, testing.data)
results_ls$RRF_train_yhat <- predict(RRF_classifier, data=training.data[,training.genes])
results_ls$RRF_classifier <- RRF_classifier
results_ls$RRF_test_yhat <- RRF_yhat
results_ls$RRF_test_ConfMat <- confusionMatrix(RRF_yhat$predictions, (Y_Test_True))
results_ls$RRF_test_ConfTab <- table(pred=RRF_yhat$predictions, truth=Y_Test_True)
results_ls$RRF_train_ConfMat <- confusionMatrix((results_ls$RRF_train_yhat$predictions), (training.data$Class))
results_ls$RRF_train_ConfTab <- table(pred=results_ls$RRF_train_yhat$predictions, truth=training.data$Class)
}
if(do_Caret_rangerRF) {
# training model with SVM
print("running Caret RF::ranger ")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
rangerRFCaret_classifier <- train(Class ~ ., data = training.data,
method = "ranger",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit)
rangerRFCaret_yhat <- predict(rangerRFCaret_classifier, testing.data)
results_ls$Caret_RFranger_train_yhat <- predict(rangerRFCaret_classifier, newdata=training.data[,training.genes])
results_ls$Caret_RFranger_classifier <- rangerRFCaret_classifier
results_ls$Caret_RFranger_test_yhat <- rangerRFCaret_yhat
results_ls$Caret_RFranger_test_ConfMat <- confusionMatrix(rangerRFCaret_yhat, (Y_Test_True))
results_ls$Caret_RFranger_test_ConfTab <- table(pred=rangerRFCaret_yhat, truth=Y_Test_True)
results_ls$Caret_RFranger_train_ConfMat <- confusionMatrix((results_ls$Caret_RFranger_train_yhat), (training.data$Class))
results_ls$Caret_RFranger_train_ConfTab <- table(pred=results_ls$Caret_RFranger_train_yhat, truth=training.data$Class)
}
if(do_Caret_pcaNNet) {
# training model with SVM
print("running Caret pcaNNet ")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
pcaNNetCaret_classifier <- train(Class ~ ., data = training.data,
method = "pcaNNet",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit)
pcaNNetCaret_yhat <- predict(pcaNNetCaret_classifier, testing.data)
results_ls$Caret_pcaNNet_train_yhat <- predict(pcaNNetCaret_classifier, newdata=training.data[,training.genes])
results_ls$Caret_pcaNNet_classifier <- pcaNNetCaret_classifier
results_ls$Caret_pcaNNet_test_yhat <- pcaNNetCaret_yhat
results_ls$Caret_pcaNNet_test_ConfMat <- confusionMatrix(pcaNNetCaret_yhat, (Y_Test_True))
results_ls$Caret_pcaNNet_test_ConfTab <- table(pred=pcaNNetCaret_yhat, truth=Y_Test_True)
results_ls$Caret_RFranger_train_ConfMat <- confusionMatrix((results_ls$Caret_pcaNNet_train_yhat), (training.data$Class))
results_ls$Caret_RFranger_train_ConfTab <- table(pred=results_ls$Caret_pcaNNet_train_yhat, truth=training.data$Class)
}
classifierNames <- names(results_ls)[grep("classifier", names(results_ls))]
results_ls$ReSamp <- resamples(lapply(classifierNames, function(clsf){
results_ls[[clsf]]
}))
results_ls$ReSampSum <- summary(results_ls$ReSamp)
bwplot(results_ls$ReSamp)
results_ls$ReSampbwplot <- recordPlot()
dotplot(results_ls$ReSamp)
results_ls$ReSampdotplot <- recordPlot()
results_ls$modelCor <- modelCor(results_ls$ReSamp)
splom(results_ls$ReSamp)
results_ls$ReSampsplom <- recordPlot()
return(results_ls)
}
eisascience/scRNASeqEisa documentation built on Aug. 7, 2019, 1:55 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Defines functions GetTraingStats GetEnsmbleVarImportance ClassifyCellsCustom
GetTraingStats <- function(ClassifierSet=NA, AvailableTrainConfMatrix="", AvailableTestConfMatrix=""){
tempDF1 <- data.frame(lapply(AvailableTrainConfMatrix, function(conftabN){
c(ClassifierSet[[conftabN]]$byClass[c(1:4, 6:10)],
ClassifierSet[[conftabN]]$overall[c(1,3,4)])
}))
colnames(tempDF1) <- AvailableTrainConfMatrix
tempDF2 <- data.frame(lapply(AvailableTestConfMatrix, function(conftabN){
c(ClassifierSet[[conftabN]]$byClass[c(1:4, 6:10)],
ClassifierSet[[conftabN]]$overall[c(1,3,4)])
}))
colnames(tempDF2) <- AvailableTestConfMatrix
data.frame(cbind(tempDF1, tempDF2))[,sort(c(AvailableTestConfMatrix, AvailableTrainConfMatrix))]
}
GetEnsmbleVarImportance <- function(ClassiferSet, AvailableClassifiers="", ScaleVarImp=T){
if(AvailableClassifiers[1]=="") AvailableClassifiers = names(ClassifierSet)
FoundVarImps <- lapply(AvailableClassifiers, function(clN){
print(clN)
tempOut <- try(varImp(ClassiferSet[[clN]],scale = ScaleVarImp), silent = T)
if(class(tempOut)[1]=="try-error") {
return(NULL)
print("No VarImp...")
} else{
tempOut2 <- tempOut$importance
if(ncol(tempOut2)<2){
tempOut2 <- data.frame(cbind(tempOut2, rep(NA, length(tempOut2))))
}
return(data.frame(t(tempOut2)))
}
})
names(FoundVarImps) <- AvailableClassifiers
FoundVarImps[sapply(FoundVarImps, is.null)] <- NULL
FoundVarImpsDF <- as.data.frame(data.table::rbindlist(FoundVarImps))
FoundVarImpsDF <- FoundVarImpsDF[is.odd(1:nrow(FoundVarImpsDF)),]
rownames(FoundVarImpsDF) <- names(FoundVarImps)
FoundVarImpsDF
}
ClassifyCellsCustom <- function(Classifier.rds.path = "", ClassifierNames="", testing.data, log10T=T, returnTraining=F){
if(file.exists(Classifier.rds.path)){
MultiClassifierResults <- readRDS(Classifier.rds.path)
Available.Classifiers <- names(MultiClassifierResults)[grepl("classifier", names(MultiClassifierResults))]
print(Available.Classifiers)
if(ClassifierNames[1]=="") ClassifierNames <- Available.Classifiers
if(length(which(ClassifierNames %in% Available.Classifiers)) < length(ClassifierNames)) {
print("Names in ClassifierNames did not match whats available")
ClassifierNames <- Available.Classifiers
}
MultiClassifier <- MultiClassifierResults[Available.Classifiers]
if(class(testing.data)[1]=="seurat") {
print("Converting Seurat Sparse Matrix to one with 0s .... ")
testing.data <- Matrix::as.matrix(t(testing.data@data))
} else {
if(!(class(testing.data)[1] %in% c("dgCMatrix", "matrix", "Matrix")) ) {
print("DGE mat not Seurat or dgCMatrix")
print("Make sure columns are cells and rows are genes and convert to expected format for speed")
testing.data <- Matrix::as.matrix((testing.data))
}
}
if(log10T) {
print("log-transforming")
testing.data <- log10(testing.data+1)
}
print("Starting Classification")
testing.data.yhat <- data.frame(lapply(Available.Classifiers, function(ClassifX){
print(ClassifX)
tempClassifier <- MultiClassifier[[ClassifX]]
y.hat <- predict(tempClassifier, newdata = testing.data)
return(y.hat)
}))
colnames(testing.data.yhat) <- Available.Classifiers
NotLevelName <- levels(testing.data.yhat[,1])[grep("Not", levels(testing.data.yhat[,1]))]
testing.data.yhat$CountNot <- apply(testing.data.yhat, 1, function(x) sum(grepl(NotLevelName, x)))
testing.data.yhat$CountTot <- rep(length(Available.Classifiers), nrow(testing.data.yhat))
testing.data.yhat$NotProb <- testing.data.yhat$CountNot/testing.data.yhat$CountTot
rownames(testing.data.yhat) <- rownames(testing.data)
return(list(yhat.DF = testing.data.yhat,
classification.levels = levels(testing.data.yhat[,1]),
Available.Classifiers = Available.Classifiers,
log10T = log10T,
DGEcolNames = colnames(testing.data),
DGErowNames = rownames(testing.data),
MultiClassifierResults = ifelse(returnTraining, MultiClassifierResults, "NULL")))
} else {
print("Check classifier path ....")
}
}
MultiClassifier_Cells_SERIII <- function (object_train,
object_test,
training.genes = NULL,
Y_Train_True = NULL,
Y_Test_True = NULL,
log10p1=T,
do_Adaboost = F,
do_Caret_LinSVM = F,
do_ranger_RF = F,
do_Caret_NaiveBayes = F,
do_Caret_NNet = F,
do_Caret_radialSVM = F,
do_ElasticNet = F,
do_tree = F,
do_tree_gini = F,
do_Caret_pcaNNet = F,
do_Caret_rangerRF = F,
do_glmStepAIC=F,
do_stepLDA=F,
crossValReps = 3,
NcrossVal = 10)
{
if(do_Adaboost) library(fastAdaboost)
if(do_Caret_LinSVM) library(caret)
if(do_Caret_NaiveBayes) library(caret)
if(do_Caret_NNet) library(caret)
if(do_ranger_RF) library(ranger)
if(do_ElasticNet) library(caret)
require(Matrix)
results_ls <- list()
training.classes <- as.vector(x = Y_Train_True)
training.genes <- SetIfNull(x = training.genes, default = rownames(x = object_train@assays$RNA@data))
training.data <- as.data.frame(x = as.matrix(x = Matrix::t(object_train@assays$RNA@data)))[,training.genes]
testing.classes <- as.vector(x = Y_Test_True)
testing.genes <- SetIfNull(x = training.genes, default = rownames(x = object_test@assays$RNA@data))
testing.data <- as.data.frame(x = as.matrix(x = Matrix::t(x = object_test@assays$RNA@data)))[,testing.genes ]
if(log10p1) training.data <- log10(training.data + 1)
if(log10p1) testing.data <- log10(testing.data + 1)
training.data$Class <- factor(x = training.classes)
#colnames(training.data) <- gsub("-", "", colnames(training.data))
#colnames(testing.data) <- gsub("-", "", colnames(testing.data))
results_ls$test_data <- testing.data
results_ls$test_Y <- Y_Test_True
results_ls$train_data <- training.data
results_ls$train_T <- Y_Train_True
#############template
# if(do_newModel){
#
#
# print("running newModel")
# TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
#
#
# newModel_classifier <- train(Class ~ ., data = training.data,
# method = "newModel",
# trControl= TrainingParameters,
# preProcess = c("scale","center"),
# na.action = na.omit)
#
#
# newModel_yhat <- predict(newModel_classifier, testing.data)
#
#
# results_ls$Caret_newModel_train_yhat <- predict(newModel_classifier, newdata=training.data[,which(colnames(training.data) !="Class")])
#
#
# results_ls$Caret_newModel_classifier <- newModel_classifier
# results_ls$Caret_newModel_test_yhat <- newModel_yhat
# results_ls$Caret_newModel_test_ConfMat <- confusionMatrix(newModel_yhat, (Y_Test_True))
# results_ls$Caret_newModel_test_ConfTab <- table(pred=newModel_yhat, truth=Y_Test_True)
#
# results_ls$Caret_newModel_train_ConfMat <- confusionMatrix((results_ls$Caret_newModel_train_yhat), (training.data$Class))
# results_ls$Caret_newModel_train_ConfTab <- table(pred=results_ls$Caret_newModel_train_yhat, truth=training.data$Class)
#
#
# colnames(training.data) <- colnames(results_ls$train_data)
# colnames(testing.data) <- colnames(results_ls$test_data)
#
#
# }
if(do_stepLDA){
print("running stepLDA")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
stepLDA_classifier <- train(Class ~ ., data = training.data,
method = "stepLDA",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit)
stepLDA_yhat <- predict(stepLDA_classifier, testing.data)
results_ls$Caret_stepLDA_train_yhat <- predict(stepLDA_classifier, newdata=training.data[,which(colnames(training.data) !="Class")])
results_ls$Caret_stepLDA_classifier <- stepLDA_classifier
results_ls$Caret_stepLDA_test_yhat <- stepLDA_yhat
results_ls$Caret_stepLDA_test_ConfMat <- confusionMatrix(stepLDA_yhat, (Y_Test_True))
results_ls$Caret_stepLDA_test_ConfTab <- table(pred=stepLDA_yhat, truth=Y_Test_True)
results_ls$Caret_stepLDA_train_ConfMat <- confusionMatrix((results_ls$Caret_stepLDA_train_yhat), (training.data$Class))
results_ls$Caret_stepLDA_train_ConfTab <- table(pred=results_ls$Caret_stepLDA_train_yhat, truth=training.data$Class)
colnames(training.data) <- colnames(results_ls$train_data)
colnames(testing.data) <- colnames(results_ls$test_data)
}
if(do_glmStepAIC){
print("running glmStepAIC")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
glmStepAIC_classifier <- train(Class ~ ., data = training.data,
method = "glmStepAIC",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit)
glmStepAIC_yhat <- predict(glmStepAIC_classifier, testing.data)
results_ls$Caret_glmStepAIC_train_yhat <- predict(glmStepAIC_classifier, newdata=training.data[,which(colnames(training.data) !="Class")])
results_ls$Caret_glmStepAIC_classifier <- glmStepAIC_classifier
results_ls$Caret_glmStepAIC_test_yhat <- glmStepAIC_yhat
results_ls$Caret_glmStepAIC_test_ConfMat <- confusionMatrix(glmStepAIC_yhat, (Y_Test_True))
results_ls$Caret_glmStepAIC_test_ConfTab <- table(pred=glmStepAIC_yhat, truth=Y_Test_True)
results_ls$Caret_glmStepAIC_train_ConfMat <- confusionMatrix((results_ls$Caret_glmStepAIC_train_yhat), (training.data$Class))
results_ls$Caret_glmStepAIC_train_ConfTab <- table(pred=results_ls$Caret_glmStepAIC_train_yhat, truth=training.data$Class)
colnames(training.data) <- colnames(results_ls$train_data)
colnames(testing.data) <- colnames(results_ls$test_data)
}
if(do_tree){
library(caret)
library(rpart)
library(rpart.plot)
print("running tree based class with rpart")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
rpart_classifier <- train(x=training.data[,setdiff(colnames(training.data), c("Class"))],
y=training.data$Class,
method = "rpart",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit,
parms = list(split = "information"),
tuneLength = 10)
prp(rpart_classifier$finalModel, box.palette = "Blues", tweak = 1.2)
results_ls$Caret_rpart_treeViz <- recordPlot()
rpart_yhat <- predict(rpart_classifier, testing.data)
results_ls$Caret_rpart_train_yhat <- predict(rpart_classifier, newdata=training.data[,which(colnames(training.data) !="Class")])
results_ls$Caret_rpart_classifier <- rpart_classifier
results_ls$Caret_rpart_test_yhat <- rpart_yhat
results_ls$Caret_rpart_test_ConfMat <- confusionMatrix(rpart_yhat, (Y_Test_True))
results_ls$Caret_rpart_test_ConfTab <- table(pred=rpart_yhat, truth=Y_Test_True)
results_ls$Caret_rpart_train_ConfMat <- confusionMatrix((results_ls$Caret_rpart_train_yhat), (training.data$Class))
results_ls$Caret_rpart_train_ConfTab <- table(pred=results_ls$Caret_rpart_train_yhat, truth=training.data$Class)
colnames(training.data) <- colnames(results_ls$train_data)
colnames(testing.data) <- colnames(results_ls$test_data)
}
if(do_tree_gini){
library(caret)
library(rpart)
library(rpart.plot)
print("running tree based class with rpart w/ gini index")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
rpartGini_classifier <- train(x=training.data[,setdiff(colnames(training.data), c("Class"))],
y=training.data$Class,
method = "rpart",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit,
parms = list(split = "gini"),
tuneLength = 10)
prp(rpartGini_classifier$finalModel, box.palette = "Blues", tweak = 1.2)
results_ls$Caret_rpartGini_treeViz <- recordPlot()
rpartGini_yhat <- predict(rpartGini_classifier, testing.data)
results_ls$Caret_rpartGini_train_yhat <- predict(rpartGini_classifier, newdata=training.data[,which(colnames(training.data) !="Class")])
results_ls$Caret_rpartGini_classifier <- rpartGini_classifier
results_ls$Caret_rpartGini_test_yhat <- rpartGini_yhat
results_ls$Caret_rpartGini_test_ConfMat <- confusionMatrix(rpartGini_yhat, (Y_Test_True))
results_ls$Caret_rpartGini_test_ConfTab <- table(pred=rpartGini_yhat, truth=Y_Test_True)
results_ls$Caret_rpartGini_train_ConfMat <- confusionMatrix((results_ls$Caret_rpartGini_train_yhat), (training.data$Class))
results_ls$Caret_rpartGini_train_ConfTab <- table(pred=results_ls$Caret_rpartGini_train_yhat, truth=training.data$Class)
results_ls
colnames(training.data) <- colnames(results_ls$train_data)
colnames(testing.data) <- colnames(results_ls$test_data)
}
if(do_ElasticNet){
library(glmnet)
print("running ElasticNet Reg.")
#print(class(training.data))
#print(head(training.data))
#print((training.data$Class))
#print(levels(training.data$Class))
#print(colnames(training.data[,setdiff(colnames(training.data), c("Class"))]))
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps,
returnResamp="all",
classProbs=TRUE)
training.data.Y <- factor(as.character(training.data$Class))
training.data.X <- training.data[,setdiff(colnames(training.data), c("Class"))]
ElasticNet_classifier <- try(caret::train(training.data.X, training.data.Y,
method = "glmnet",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit), silent = T)
#print(class(ElasticNet_classifier))
if(class(ElasticNet_classifier)[1]=="try-error") {
print("first model failed, tring with alternate alpha and lambda...")
ElasticNet_classifier <- try(caret::train(training.data.X, training.data.Y,
method = "glmnet",
metric = "ROC",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit,
tuneGrid = expand.grid(alpha = c(0.1, 1),
lambda = c(0.001, 0.0001))),
silent = T)
}
if(class(ElasticNet_classifier)[1]=="try-error") {
warning("ElasticNet classifier failed ")
} else {
ElasticNet_yhat <- predict(ElasticNet_classifier, testing.data)
results_ls$Caret_ElasticNet_train_yhat <- predict(ElasticNet_classifier, newdata=training.data[,training.genes])
results_ls$Caret_ElasticNet_classifier <- ElasticNet_classifier
results_ls$Caret_ElasticNet_test_yhat <- ElasticNet_yhat
results_ls$Caret_ElasticNet_test_ConfMat <- confusionMatrix(ElasticNet_yhat, (Y_Test_True))
results_ls$Caret_ElasticNet_test_ConfTab <- table(pred=ElasticNet_yhat, truth=Y_Test_True)
results_ls$Caret_ElasticNet_train_ConfMat <- confusionMatrix((results_ls$Caret_ElasticNet_train_yhat), (training.data$Class))
results_ls$Caret_ElasticNet_train_ConfTab <- table(pred=results_ls$Caret_ElasticNet_train_yhat, truth=training.data$Class)
}
}
if(do_Adaboost){
print("running Adaboost")
Adaboost_classifier <- fastAdaboost::adaboost(Class ~ ., training.data, NcrossVal)
y_hat_Adaboost <- predict(Adaboost_classifier, newdata=testing.data)
results_ls$Adaboost_train_yhat <- predict(Adaboost_classifier, newdata=training.data[,training.genes])
results_ls$Adaboost_classifier <- Adaboost_classifier
results_ls$Adaboost_test_yhat <- y_hat_Adaboost
results_ls$Adaboost_test_ConfMat <- confusionMatrix((y_hat_Adaboost$class), (Y_Test_True))
results_ls$Adaboost_test_ConfTab <- table(pred=y_hat_Adaboost$class, truth=Y_Test_True)
results_ls$Adaboost_train_ConfMat <- confusionMatrix((results_ls$Adaboost_train_yhat$class), (training.data$Class))
results_ls$Adaboost_train_ConfTab <- table(pred=results_ls$Adaboost_train_yhat$class, truth=training.data$Class)
}
if(do_Caret_LinSVM) {
# training model with SVM
print("running Caret LinSVM")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
SVM_linear_classifier <- train(Class ~ ., data = training.data,
method = "svmLinear",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit)
SVM_Lin_yhat <- predict(SVM_linear_classifier, testing.data)
results_ls$Caret_LinSVM_train_yhat <- predict(SVM_linear_classifier, newdata=training.data[,training.genes])
results_ls$Caret_LinSVM_classifier <- SVM_linear_classifier
results_ls$Caret_LinSVM_test_yhat <- SVM_Lin_yhat
results_ls$Caret_LinSVM_test_ConfMat <- confusionMatrix(SVM_Lin_yhat, (Y_Test_True))
results_ls$Caret_LinSVM_test_ConfTab <- table(pred=SVM_Lin_yhat, truth=Y_Test_True)
results_ls$Caret_LinSVM_train_ConfMat <- confusionMatrix((results_ls$Caret_LinSVM_train_yhat), (training.data$Class))
results_ls$Caret_LinSVM_train_ConfTab <- table(pred=results_ls$Caret_LinSVM_train_yhat, truth=training.data$Class)
}
if(do_Caret_radialSVM) {
# training model with SVM
print("running Caret radialSVM")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
SVM_radial_classifier <- train(Class ~ ., data = training.data,
method = "svmRadial",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit)
SVM_Radial_yhat <- predict(SVM_radial_classifier, testing.data)
results_ls$Caret_radialSVM_train_yhat <- predict(SVM_radial_classifier, newdata=training.data[,training.genes])
results_ls$Caret_radialSVM_classifier <- SVM_radial_classifier
results_ls$Caret_radialSVM_test_yhat <- SVM_Radial_yhat
results_ls$Caret_radialSVM_test_ConfMat <- confusionMatrix(SVM_Radial_yhat, (Y_Test_True))
results_ls$Caret_radialSVM_test_ConfTab <- table(pred=SVM_Radial_yhat, truth=Y_Test_True)
results_ls$Caret_radialSVM_train_ConfMat <- confusionMatrix((results_ls$Caret_radialSVM_train_yhat), (training.data$Class))
results_ls$Caret_radialSVM_train_ConfTab <- table(pred=results_ls$Caret_radialSVM_train_yhat, truth=training.data$Class)
}
if(do_Caret_NaiveBayes) {
print("running Naive Bayes")
# training model with SVM
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
#Naive Bayes algorithm
NaiveBayes_classifier <- train(Class ~ ., data = training.data,
method = "nb",
preProcess=c("scale","center"),
trControl= TrainingParameters,
na.action = na.omit, silent = T)
NaiveBayes_yhat <- predict(NaiveBayes_classifier, testing.data)
results_ls$NaiveBayes_train_yhat <- predict(NaiveBayes_classifier, newdata=training.data[,training.genes])
results_ls$NaiveBayes_classifier <- NaiveBayes_classifier
results_ls$NaiveBayes_test_yhat <- NaiveBayes_yhat
results_ls$NaiveBayes_test_ConfMat <- confusionMatrix(NaiveBayes_yhat, (Y_Test_True))
results_ls$NaiveBayes_test_ConfTab <- table(pred=NaiveBayes_yhat, truth=Y_Test_True)
results_ls$NaiveBayes_train_ConfMat <- confusionMatrix((results_ls$NaiveBayes_train_yhat), (training.data$Class))
results_ls$NaiveBayes_train_ConfTab <- table(pred=results_ls$NaiveBayes_train_yhat, truth=training.data$Class)
}
if(do_Caret_NNet) {
print("running NNet")
# training model with SVM
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
#Naive Bayes algorithm
NNet_classifier <- train(Class ~ ., data = training.data,
method = "nnet",
preProcess=c("scale","center"),
trControl= TrainingParameters,
na.action = na.omit)
NNet_yhat <- predict(NNet_classifier, testing.data)
results_ls$NNet_train_yhat <- predict(NNet_classifier, newdata=training.data[,training.genes])
results_ls$NNet_classifier <- NNet_classifier
results_ls$NNet_test_yhat <- NNet_yhat
results_ls$NNet_test_ConfMat <- confusionMatrix(NNet_yhat, (Y_Test_True))
results_ls$NNet_test_ConfTab <- table(pred=NNet_yhat, truth=Y_Test_True)
results_ls$NNet_train_ConfMat <- confusionMatrix((results_ls$NNet_train_yhat), (training.data$Class))
results_ls$NNet_train_ConfTab <- table(pred=results_ls$NNet_train_yhat, truth=training.data$Class)
}
if(do_ranger_RF) {
# training
print("running ranger-RF")
RRF_classifier <- ranger::ranger(data = training.data,
dependent.variable.name = "Class",
classification = TRUE,
write.forest = TRUE)
RRF_yhat <- predict(RRF_classifier, testing.data)
results_ls$RRF_train_yhat <- predict(RRF_classifier, data=training.data[,training.genes])
results_ls$RRF_classifier <- RRF_classifier
results_ls$RRF_test_yhat <- RRF_yhat
results_ls$RRF_test_ConfMat <- confusionMatrix(RRF_yhat$predictions, (Y_Test_True))
results_ls$RRF_test_ConfTab <- table(pred=RRF_yhat$predictions, truth=Y_Test_True)
results_ls$RRF_train_ConfMat <- confusionMatrix((results_ls$RRF_train_yhat$predictions), (training.data$Class))
results_ls$RRF_train_ConfTab <- table(pred=results_ls$RRF_train_yhat$predictions, truth=training.data$Class)
}
if(do_Caret_rangerRF) {
# training model with SVM
print("running Caret RF::ranger ")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
rangerRFCaret_classifier <- train(Class ~ ., data = training.data,
method = "ranger",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit)
rangerRFCaret_yhat <- predict(rangerRFCaret_classifier, testing.data)
results_ls$Caret_RFranger_train_yhat <- predict(rangerRFCaret_classifier, newdata=training.data[,training.genes])
results_ls$Caret_RFranger_classifier <- rangerRFCaret_classifier
results_ls$Caret_RFranger_test_yhat <- rangerRFCaret_yhat
results_ls$Caret_RFranger_test_ConfMat <- confusionMatrix(rangerRFCaret_yhat, (Y_Test_True))
results_ls$Caret_RFranger_test_ConfTab <- table(pred=rangerRFCaret_yhat, truth=Y_Test_True)
results_ls$Caret_RFranger_train_ConfMat <- confusionMatrix((results_ls$Caret_RFranger_train_yhat), (training.data$Class))
results_ls$Caret_RFranger_train_ConfTab <- table(pred=results_ls$Caret_RFranger_train_yhat, truth=training.data$Class)
}
if(do_Caret_pcaNNet) {
# training model with SVM
print("running Caret pcaNNet ")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
pcaNNetCaret_classifier <- train(Class ~ ., data = training.data,
method = "pcaNNet",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit)
pcaNNetCaret_yhat <- predict(pcaNNetCaret_classifier, testing.data)
results_ls$Caret_pcaNNet_train_yhat <- predict(pcaNNetCaret_classifier, newdata=training.data[,training.genes])
results_ls$Caret_pcaNNet_classifier <- pcaNNetCaret_classifier
results_ls$Caret_pcaNNet_test_yhat <- pcaNNetCaret_yhat
results_ls$Caret_pcaNNet_test_ConfMat <- confusionMatrix(pcaNNetCaret_yhat, (Y_Test_True))
results_ls$Caret_pcaNNet_test_ConfTab <- table(pred=pcaNNetCaret_yhat, truth=Y_Test_True)
results_ls$Caret_RFranger_train_ConfMat <- confusionMatrix((results_ls$Caret_pcaNNet_train_yhat), (training.data$Class))
results_ls$Caret_RFranger_train_ConfTab <- table(pred=results_ls$Caret_pcaNNet_train_yhat, truth=training.data$Class)
}
classifierNames <- names(results_ls)[grep("classifier", names(results_ls))]
results_ls$ReSamp <- resamples(lapply(classifierNames, function(clsf){
results_ls[[clsf]]
}))
results_ls$ReSampSum <- summary(results_ls$ReSamp)
bwplot(results_ls$ReSamp)
results_ls$ReSampbwplot <- recordPlot()
dotplot(results_ls$ReSamp)
results_ls$ReSampdotplot <- recordPlot()
results_ls$modelCor <- modelCor(results_ls$ReSamp)
splom(results_ls$ReSamp)
results_ls$ReSampsplom <- recordPlot()
return(results_ls)
}
MultiClassifier_Cells <- function (object_train,
object_test,
training.genes = NULL,
Y_Train_True = NULL,
Y_Test_True = NULL,
log10p1=T,
do_Adaboost = F,
do_Caret_LinSVM = F,
do_ranger_RF = F,
do_Caret_NaiveBayes = F,
do_Caret_NNet = F,
do_Caret_radialSVM = F,
do_ElasticNet = F,
do_tree = F,
do_tree_gini = F,
do_Caret_pcaNNet = F,
do_Caret_rangerRF = F,
do_glmStepAIC=F,
do_stepLDA=F,
crossValReps = 3,
NcrossVal = 10)
{
if(do_Adaboost) library(fastAdaboost)
if(do_Caret_LinSVM) library(caret)
if(do_Caret_NaiveBayes) library(caret)
if(do_Caret_NNet) library(caret)
if(do_ranger_RF) library(ranger)
if(do_ElasticNet) library(caret)
require(Matrix)
results_ls <- list()
training.classes <- as.vector(x = Y_Train_True)
training.genes <- SetIfNull(x = training.genes, default = rownames(x = object_train@assays$RNA@data))
training.data <- as.data.frame(x = as.matrix(x = Matrix::t(object_train@assays$RNA@data)))[,training.genes]
testing.classes <- as.vector(x = Y_Test_True)
testing.genes <- SetIfNull(x = training.genes, default = rownames(x = object_test@assays$RNA@data))
testing.data <- as.data.frame(x = as.matrix(x = Matrix::t(x = object_test@assays$RNA@data)))[,testing.genes ]
if(log10p1) training.data <- log10(training.data + 1)
if(log10p1) testing.data <- log10(testing.data + 1)
training.data$Class <- factor(x = training.classes)
#colnames(training.data) <- gsub("-", "", colnames(training.data))
#colnames(testing.data) <- gsub("-", "", colnames(testing.data))
results_ls$test_data <- testing.data
results_ls$test_Y <- Y_Test_True
results_ls$train_data <- training.data
results_ls$train_T <- Y_Train_True
#############template
# if(do_newModel){
#
#
# print("running newModel")
# TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
#
#
# newModel_classifier <- train(Class ~ ., data = training.data,
# method = "newModel",
# trControl= TrainingParameters,
# preProcess = c("scale","center"),
# na.action = na.omit)
#
#
# newModel_yhat <- predict(newModel_classifier, testing.data)
#
#
# results_ls$Caret_newModel_train_yhat <- predict(newModel_classifier, newdata=training.data[,which(colnames(training.data) !="Class")])
#
#
# results_ls$Caret_newModel_classifier <- newModel_classifier
# results_ls$Caret_newModel_test_yhat <- newModel_yhat
# results_ls$Caret_newModel_test_ConfMat <- confusionMatrix(newModel_yhat, (Y_Test_True))
# results_ls$Caret_newModel_test_ConfTab <- table(pred=newModel_yhat, truth=Y_Test_True)
#
# results_ls$Caret_newModel_train_ConfMat <- confusionMatrix((results_ls$Caret_newModel_train_yhat), (training.data$Class))
# results_ls$Caret_newModel_train_ConfTab <- table(pred=results_ls$Caret_newModel_train_yhat, truth=training.data$Class)
#
#
# colnames(training.data) <- colnames(results_ls$train_data)
# colnames(testing.data) <- colnames(results_ls$test_data)
#
#
# }
if(do_stepLDA){
print("running stepLDA")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
stepLDA_classifier <- train(Class ~ ., data = training.data,
method = "stepLDA",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit)
stepLDA_yhat <- predict(stepLDA_classifier, testing.data)
results_ls$Caret_stepLDA_train_yhat <- predict(stepLDA_classifier, newdata=training.data[,which(colnames(training.data) !="Class")])
results_ls$Caret_stepLDA_classifier <- stepLDA_classifier
results_ls$Caret_stepLDA_test_yhat <- stepLDA_yhat
results_ls$Caret_stepLDA_test_ConfMat <- confusionMatrix(stepLDA_yhat, (Y_Test_True))
results_ls$Caret_stepLDA_test_ConfTab <- table(pred=stepLDA_yhat, truth=Y_Test_True)
results_ls$Caret_stepLDA_train_ConfMat <- confusionMatrix((results_ls$Caret_stepLDA_train_yhat), (training.data$Class))
results_ls$Caret_stepLDA_train_ConfTab <- table(pred=results_ls$Caret_stepLDA_train_yhat, truth=training.data$Class)
colnames(training.data) <- colnames(results_ls$train_data)
colnames(testing.data) <- colnames(results_ls$test_data)
}
if(do_glmStepAIC){
print("running glmStepAIC")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
glmStepAIC_classifier <- train(Class ~ ., data = training.data,
method = "glmStepAIC",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit)
glmStepAIC_yhat <- predict(glmStepAIC_classifier, testing.data)
results_ls$Caret_glmStepAIC_train_yhat <- predict(glmStepAIC_classifier, newdata=training.data[,which(colnames(training.data) !="Class")])
results_ls$Caret_glmStepAIC_classifier <- glmStepAIC_classifier
results_ls$Caret_glmStepAIC_test_yhat <- glmStepAIC_yhat
results_ls$Caret_glmStepAIC_test_ConfMat <- confusionMatrix(glmStepAIC_yhat, (Y_Test_True))
results_ls$Caret_glmStepAIC_test_ConfTab <- table(pred=glmStepAIC_yhat, truth=Y_Test_True)
results_ls$Caret_glmStepAIC_train_ConfMat <- confusionMatrix((results_ls$Caret_glmStepAIC_train_yhat), (training.data$Class))
results_ls$Caret_glmStepAIC_train_ConfTab <- table(pred=results_ls$Caret_glmStepAIC_train_yhat, truth=training.data$Class)
colnames(training.data) <- colnames(results_ls$train_data)
colnames(testing.data) <- colnames(results_ls$test_data)
}
if(do_tree){
library(caret)
library(rpart)
library(rpart.plot)
print("running tree based class with rpart")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
rpart_classifier <- train(x=training.data[,setdiff(colnames(training.data), c("Class"))],
y=training.data$Class,
method = "rpart",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit,
parms = list(split = "information"),
tuneLength = 10)
prp(rpart_classifier$finalModel, box.palette = "Blues", tweak = 1.2)
results_ls$Caret_rpart_treeViz <- recordPlot()
rpart_yhat <- predict(rpart_classifier, testing.data)
results_ls$Caret_rpart_train_yhat <- predict(rpart_classifier, newdata=training.data[,which(colnames(training.data) !="Class")])
results_ls$Caret_rpart_classifier <- rpart_classifier
results_ls$Caret_rpart_test_yhat <- rpart_yhat
results_ls$Caret_rpart_test_ConfMat <- confusionMatrix(rpart_yhat, (Y_Test_True))
results_ls$Caret_rpart_test_ConfTab <- table(pred=rpart_yhat, truth=Y_Test_True)
results_ls$Caret_rpart_train_ConfMat <- confusionMatrix((results_ls$Caret_rpart_train_yhat), (training.data$Class))
results_ls$Caret_rpart_train_ConfTab <- table(pred=results_ls$Caret_rpart_train_yhat, truth=training.data$Class)
colnames(training.data) <- colnames(results_ls$train_data)
colnames(testing.data) <- colnames(results_ls$test_data)
}
if(do_tree_gini){
library(caret)
library(rpart)
library(rpart.plot)
print("running tree based class with rpart w/ gini index")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
rpartGini_classifier <- train(x=training.data[,setdiff(colnames(training.data), c("Class"))],
y=training.data$Class,
method = "rpart",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit,
parms = list(split = "gini"),
tuneLength = 10)
prp(rpartGini_classifier$finalModel, box.palette = "Blues", tweak = 1.2)
results_ls$Caret_rpartGini_treeViz <- recordPlot()
rpartGini_yhat <- predict(rpartGini_classifier, testing.data)
results_ls$Caret_rpartGini_train_yhat <- predict(rpartGini_classifier, newdata=training.data[,which(colnames(training.data) !="Class")])
results_ls$Caret_rpartGini_classifier <- rpartGini_classifier
results_ls$Caret_rpartGini_test_yhat <- rpartGini_yhat
results_ls$Caret_rpartGini_test_ConfMat <- confusionMatrix(rpartGini_yhat, (Y_Test_True))
results_ls$Caret_rpartGini_test_ConfTab <- table(pred=rpartGini_yhat, truth=Y_Test_True)
results_ls$Caret_rpartGini_train_ConfMat <- confusionMatrix((results_ls$Caret_rpartGini_train_yhat), (training.data$Class))
results_ls$Caret_rpartGini_train_ConfTab <- table(pred=results_ls$Caret_rpartGini_train_yhat, truth=training.data$Class)
results_ls
colnames(training.data) <- colnames(results_ls$train_data)
colnames(testing.data) <- colnames(results_ls$test_data)
}
if(do_ElasticNet){
library(glmnet)
print("running ElasticNet Reg.")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
ElasticNet_classifier <- try(train(x=training.data[,setdiff(colnames(training.data), c("Class"))],
y=training.data$Class,
method = "glmnet",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit), silent = T)
if(class(ElasticNet_classifier)[1]=="try-error") {
ElasticNet_classifier <- try(train(x=training.data[,setdiff(colnames(training.data), c("Class"))],
y=training.data$Class,
method = "glmnet",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit,
tuneGrid = expand.grid(alpha = 0.1, lambda =0.01)),
silent = T)
}
ElasticNet_yhat <- predict(ElasticNet_classifier, testing.data)
results_ls$Caret_ElasticNet_train_yhat <- predict(ElasticNet_classifier, newdata=training.data[,training.genes])
results_ls$Caret_ElasticNet_classifier <- ElasticNet_classifier
results_ls$Caret_ElasticNet_test_yhat <- ElasticNet_yhat
results_ls$Caret_ElasticNet_test_ConfMat <- confusionMatrix(ElasticNet_yhat, (Y_Test_True))
results_ls$Caret_ElasticNet_test_ConfTab <- table(pred=ElasticNet_yhat, truth=Y_Test_True)
results_ls$Caret_ElasticNet_train_ConfMat <- confusionMatrix((results_ls$Caret_ElasticNet_train_yhat), (training.data$Class))
results_ls$Caret_ElasticNet_train_ConfTab <- table(pred=results_ls$Caret_ElasticNet_train_yhat, truth=training.data$Class)
}
if(do_Adaboost){
print("running Adaboost")
Adaboost_classifier <- fastAdaboost::adaboost(Class ~ ., training.data, NcrossVal)
y_hat_Adaboost <- predict(Adaboost_classifier, newdata=testing.data)
results_ls$Adaboost_train_yhat <- predict(Adaboost_classifier, newdata=training.data[,training.genes])
results_ls$Adaboost_classifier <- Adaboost_classifier
results_ls$Adaboost_test_yhat <- y_hat_Adaboost
results_ls$Adaboost_test_ConfMat <- confusionMatrix((y_hat_Adaboost$class), (Y_Test_True))
results_ls$Adaboost_test_ConfTab <- table(pred=y_hat_Adaboost$class, truth=Y_Test_True)
results_ls$Adaboost_train_ConfMat <- confusionMatrix((results_ls$Adaboost_train_yhat$class), (training.data$Class))
results_ls$Adaboost_train_ConfTab <- table(pred=results_ls$Adaboost_train_yhat$class, truth=training.data$Class)
}
if(do_Caret_LinSVM) {
# training model with SVM
print("running Caret LinSVM")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
SVM_linear_classifier <- train(Class ~ ., data = training.data,
method = "svmLinear",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit)
SVM_Lin_yhat <- predict(SVM_linear_classifier, testing.data)
results_ls$Caret_LinSVM_train_yhat <- predict(SVM_linear_classifier, newdata=training.data[,training.genes])
results_ls$Caret_LinSVM_classifier <- SVM_linear_classifier
results_ls$Caret_LinSVM_test_yhat <- SVM_Lin_yhat
results_ls$Caret_LinSVM_test_ConfMat <- confusionMatrix(SVM_Lin_yhat, (Y_Test_True))
results_ls$Caret_LinSVM_test_ConfTab <- table(pred=SVM_Lin_yhat, truth=Y_Test_True)
results_ls$Caret_LinSVM_train_ConfMat <- confusionMatrix((results_ls$Caret_LinSVM_train_yhat), (training.data$Class))
results_ls$Caret_LinSVM_train_ConfTab <- table(pred=results_ls$Caret_LinSVM_train_yhat, truth=training.data$Class)
}
if(do_Caret_radialSVM) {
# training model with SVM
print("running Caret radialSVM")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
SVM_radial_classifier <- train(Class ~ ., data = training.data,
method = "svmRadial",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit)
SVM_Radial_yhat <- predict(SVM_radial_classifier, testing.data)
results_ls$Caret_radialSVM_train_yhat <- predict(SVM_radial_classifier, newdata=training.data[,training.genes])
results_ls$Caret_radialSVM_classifier <- SVM_radial_classifier
results_ls$Caret_radialSVM_test_yhat <- SVM_Radial_yhat
results_ls$Caret_radialSVM_test_ConfMat <- confusionMatrix(SVM_Radial_yhat, (Y_Test_True))
results_ls$Caret_radialSVM_test_ConfTab <- table(pred=SVM_Radial_yhat, truth=Y_Test_True)
results_ls$Caret_radialSVM_train_ConfMat <- confusionMatrix((results_ls$Caret_radialSVM_train_yhat), (training.data$Class))
results_ls$Caret_radialSVM_train_ConfTab <- table(pred=results_ls$Caret_radialSVM_train_yhat, truth=training.data$Class)
}
if(do_Caret_NaiveBayes) {
print("running Naive Bayes")
# training model with SVM
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
#Naive Bayes algorithm
NaiveBayes_classifier <- train(Class ~ ., data = training.data,
method = "nb",
preProcess=c("scale","center"),
trControl= TrainingParameters,
na.action = na.omit, silent = T)
NaiveBayes_yhat <- predict(NaiveBayes_classifier, testing.data)
results_ls$NaiveBayes_train_yhat <- predict(NaiveBayes_classifier, newdata=training.data[,training.genes])
results_ls$NaiveBayes_classifier <- NaiveBayes_classifier
results_ls$NaiveBayes_test_yhat <- NaiveBayes_yhat
results_ls$NaiveBayes_test_ConfMat <- confusionMatrix(NaiveBayes_yhat, (Y_Test_True))
results_ls$NaiveBayes_test_ConfTab <- table(pred=NaiveBayes_yhat, truth=Y_Test_True)
results_ls$NaiveBayes_train_ConfMat <- confusionMatrix((results_ls$NaiveBayes_train_yhat), (training.data$Class))
results_ls$NaiveBayes_train_ConfTab <- table(pred=results_ls$NaiveBayes_train_yhat, truth=training.data$Class)
}
if(do_Caret_NNet) {
print("running NNet")
# training model with SVM
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
#Naive Bayes algorithm
NNet_classifier <- train(Class ~ ., data = training.data,
method = "nnet",
preProcess=c("scale","center"),
trControl= TrainingParameters,
na.action = na.omit)
NNet_yhat <- predict(NNet_classifier, testing.data)
results_ls$NNet_train_yhat <- predict(NNet_classifier, newdata=training.data[,training.genes])
results_ls$NNet_classifier <- NNet_classifier
results_ls$NNet_test_yhat <- NNet_yhat
results_ls$NNet_test_ConfMat <- confusionMatrix(NNet_yhat, (Y_Test_True))
results_ls$NNet_test_ConfTab <- table(pred=NNet_yhat, truth=Y_Test_True)
results_ls$NNet_train_ConfMat <- confusionMatrix((results_ls$NNet_train_yhat), (training.data$Class))
results_ls$NNet_train_ConfTab <- table(pred=results_ls$NNet_train_yhat, truth=training.data$Class)
}
if(do_ranger_RF) {
# training
print("running ranger-RF")
RRF_classifier <- ranger::ranger(data = training.data,
dependent.variable.name = "Class",
classification = TRUE,
write.forest = TRUE)
RRF_yhat <- predict(RRF_classifier, testing.data)
results_ls$RRF_train_yhat <- predict(RRF_classifier, data=training.data[,training.genes])
results_ls$RRF_classifier <- RRF_classifier
results_ls$RRF_test_yhat <- RRF_yhat
results_ls$RRF_test_ConfMat <- confusionMatrix(RRF_yhat$predictions, (Y_Test_True))
results_ls$RRF_test_ConfTab <- table(pred=RRF_yhat$predictions, truth=Y_Test_True)
results_ls$RRF_train_ConfMat <- confusionMatrix((results_ls$RRF_train_yhat$predictions), (training.data$Class))
results_ls$RRF_train_ConfTab <- table(pred=results_ls$RRF_train_yhat$predictions, truth=training.data$Class)
}
if(do_Caret_rangerRF) {
# training model with SVM
print("running Caret RF::ranger ")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
rangerRFCaret_classifier <- train(Class ~ ., data = training.data,
method = "ranger",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit)
rangerRFCaret_yhat <- predict(rangerRFCaret_classifier, testing.data)
results_ls$Caret_RFranger_train_yhat <- predict(rangerRFCaret_classifier, newdata=training.data[,training.genes])
results_ls$Caret_RFranger_classifier <- rangerRFCaret_classifier
results_ls$Caret_RFranger_test_yhat <- rangerRFCaret_yhat
results_ls$Caret_RFranger_test_ConfMat <- confusionMatrix(rangerRFCaret_yhat, (Y_Test_True))
results_ls$Caret_RFranger_test_ConfTab <- table(pred=rangerRFCaret_yhat, truth=Y_Test_True)
results_ls$Caret_RFranger_train_ConfMat <- confusionMatrix((results_ls$Caret_RFranger_train_yhat), (training.data$Class))
results_ls$Caret_RFranger_train_ConfTab <- table(pred=results_ls$Caret_RFranger_train_yhat, truth=training.data$Class)
}
if(do_Caret_pcaNNet) {
# training model with SVM
print("running Caret pcaNNet ")
TrainingParameters <- trainControl(method = "repeatedcv", number = NcrossVal, repeats=crossValReps)
pcaNNetCaret_classifier <- train(Class ~ ., data = training.data,
method = "pcaNNet",
trControl= TrainingParameters,
preProcess = c("scale","center"),
na.action = na.omit)
pcaNNetCaret_yhat <- predict(pcaNNetCaret_classifier, testing.data)
results_ls$Caret_pcaNNet_train_yhat <- predict(pcaNNetCaret_classifier, newdata=training.data[,training.genes])
results_ls$Caret_pcaNNet_classifier <- pcaNNetCaret_classifier
results_ls$Caret_pcaNNet_test_yhat <- pcaNNetCaret_yhat
results_ls$Caret_pcaNNet_test_ConfMat <- confusionMatrix(pcaNNetCaret_yhat, (Y_Test_True))
results_ls$Caret_pcaNNet_test_ConfTab <- table(pred=pcaNNetCaret_yhat, truth=Y_Test_True)
results_ls$Caret_RFranger_train_ConfMat <- confusionMatrix((results_ls$Caret_pcaNNet_train_yhat), (training.data$Class))
results_ls$Caret_RFranger_train_ConfTab <- table(pred=results_ls$Caret_pcaNNet_train_yhat, truth=training.data$Class)
}
classifierNames <- names(results_ls)[grep("classifier", names(results_ls))]
results_ls$ReSamp <- resamples(lapply(classifierNames, function(clsf){
results_ls[[clsf]]
}))
results_ls$ReSampSum <- summary(results_ls$ReSamp)
bwplot(results_ls$ReSamp)
results_ls$ReSampbwplot <- recordPlot()
dotplot(results_ls$ReSamp)
results_ls$ReSampdotplot <- recordPlot()
results_ls$modelCor <- modelCor(results_ls$ReSamp)
splom(results_ls$ReSamp)
results_ls$ReSampsplom <- recordPlot()
return(results_ls)
}
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.