Nothing
R/RegresionBenchmark.R
In FRESA.CAD: Feature Selection Algorithms for Computer Aided Diagnosis
Defines functions
RegresionBenchmark
Documented in
RegresionBenchmark
RegresionBenchmark <- function(theData = NULL, theOutcome = "Class", reps = 100, trainFraction = 0.5,referenceCV = NULL,referenceName = "Reference",referenceFilterName="Reference")
{
if (!requireNamespace("e1071", quietly = TRUE)) {
install.packages("e1071", dependencies = TRUE)
}
if (!requireNamespace("randomForest", quietly = TRUE)) {
install.packages("randomForest", dependencies = TRUE)
}
if (!requireNamespace("rpart", quietly = TRUE)) {
install.packages("rpart", dependencies = TRUE)
}
if (!requireNamespace("MASS", quietly = TRUE)) {
install.packages("MASS", dependencies = TRUE)
}
if (is.null(theData))
{
if (exists("theDataSet", envir=FRESAcacheEnv))
{
theData <- get("theDataSet", envir=FRESAcacheEnv);
theOutcome <- get("theDataOutcome", envir=FRESAcacheEnv);
}
}
else
{
assign("theDataSet",theData,FRESAcacheEnv);
assign("theDataOutcome",theOutcome,FRESAcacheEnv);
}
RMSETable <- NULL
CorTable <- NULL
BiasTable <- NULL
MAETable <- NULL
CorSpearman <- NULL
RMSETable_filter <- NULL
CorTable_filter <- NULL
BiasTable_filter <- NULL
MAETable_filter <- NULL
CorSpearman_filter <- NULL
fmeth_0 <- NULL;
# par(mfrow = c(1,1));
FilterMethod <- function(regresionfun = e1071::svm, regnamefunc = "",...)
{
RMSETable_f <- NULL
CorTable_f <- NULL
MAETable_f <- NULL
CorSpearman_f<- NULL
BiasTable_f <- NULL
parm <- list(...);
rcvFilter_Reference <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = referenceCV$selectedFeaturesSet,...);
stats <- predictionStats_regression(rcvFilter_Reference$medianTest);
CorTable_f <- rbind(CorTable_f,stats$corci);
BiasTable_f <- rbind(BiasTable_f,stats$biasci);
RMSETable_f <- rbind(RMSETable_f,stats$RMSEci);
CorSpearman_f <- rbind(CorSpearman_f,stats$spearmanci);
MAETable_f <- rbind(MAETable_f,stats$MAEci);
rcvFilter_LASSO <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = rcvLASSO$selectedFeaturesSet,...);
stats <- predictionStats_regression(rcvFilter_LASSO$medianTest);
CorTable_f <- rbind(CorTable_f,stats$corci);
BiasTable_f <- rbind(BiasTable_f,stats$biasci);
RMSETable_f <- rbind(RMSETable_f,stats$RMSEci);
CorSpearman_f <- rbind(CorSpearman_f,stats$spearmanci);
MAETable_f <- rbind(MAETable_f,stats$MAEci);
rcvFilter_RPART <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = rcvRPART$selectedFeaturesSet,...);
stats <- predictionStats_regression(rcvFilter_RPART$medianTest);
CorTable_f <- rbind(CorTable_f,stats$corci);
BiasTable_f <- rbind(BiasTable_f,stats$biasci);
RMSETable_f <- rbind(RMSETable_f,stats$RMSEci);
CorSpearman_f <- rbind(CorSpearman_f,stats$spearmanci);
MAETable_f <- rbind(MAETable_f,stats$MAEci);
selectedFeaturesSet <- rcvRF$selectedFeaturesSet
for (i in 1:length(selectedFeaturesSet))
{
if (length(referenceCV$selectedFeaturesSet[[i]]) > 1)
{
if (length(selectedFeaturesSet[[i]]) > length(referenceCV$selectedFeaturesSet[[i]]))
{
selectedFeaturesSet[[i]] <- selectedFeaturesSet[[i]][1:length(referenceCV$selectedFeaturesSet[[i]])];
}
}
else # the top five or RF
{
warning ("Less than 2 features, then will keep the top five of RF\n")
if (length(selectedFeaturesSet[[i]]) > 5)
{
selectedFeaturesSet[[i]] <- selectedFeaturesSet[[i]][1:5];
}
}
}
rcvFilter_RF <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = selectedFeaturesSet,...);
stats <- predictionStats_regression(rcvFilter_RF$medianTest);
CorTable_f <- rbind(CorTable_f,stats$corci);
BiasTable_f <- rbind(BiasTable_f,stats$biasci);
RMSETable_f <- rbind(RMSETable_f,stats$RMSEci);
CorSpearman_f <- rbind(CorSpearman_f,stats$spearmanci);
MAETable_f <- rbind(MAETable_f,stats$MAEci);
if (is.null(fmeth_0))
{
rcvFilter_FT <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = univariate_residual,featureSelection.control = list(uniTest = "Ftest",limit = 0.35,thr = 0.975),...);
}
else
{
rcvFilter_FT <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = fmeth_0$rcvFilter_FT$selectedFeaturesSet,...);
}
stats <- predictionStats_regression(rcvFilter_FT$medianTest);
CorTable_f <- rbind(CorTable_f,stats$corci);
BiasTable_f <- rbind(BiasTable_f,stats$biasci);
RMSETable_f <- rbind(RMSETable_f,stats$RMSEci);
CorSpearman_f <- rbind(CorSpearman_f,stats$spearmanci);
MAETable_f <- rbind(MAETable_f,stats$MAEci);
if (is.null(fmeth_0))
{
rcvFilter_Wt <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = univariate_residual,featureSelection.control = list(uniTest = "Wilcox",limit = 0.35,thr = 0.975),...);
}
else
{
rcvFilter_Wt <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = fmeth_0$rcvFilter_Wt$selectedFeaturesSet,...);
}
stats <- predictionStats_regression(rcvFilter_Wt$medianTest);
CorTable_f <- rbind(CorTable_f,stats$corci);
BiasTable_f <- rbind(BiasTable_f,stats$biasci);
RMSETable_f <- rbind(RMSETable_f,stats$RMSEci);
CorSpearman_f <- rbind(CorSpearman_f,stats$spearmanci);
MAETable_f <- rbind(MAETable_f,stats$MAEci);
if (is.null(fmeth_0))
{
rcvFilter_pearson <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = univariate_correlation,featureSelection.control = list(method = "pearson",limit = 0.35,thr = 0.975),...);
}
else
{
rcvFilter_pearson <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = fmeth_0$rcvFilter_pearson$selectedFeaturesSet,...);
}
stats <- predictionStats_regression(rcvFilter_pearson$medianTest);
CorTable_f <- rbind(CorTable_f,stats$corci);
BiasTable_f <- rbind(BiasTable_f,stats$biasci);
RMSETable_f <- rbind(RMSETable_f,stats$RMSEci);
CorSpearman_f <- rbind(CorSpearman_f,stats$spearmanci);
MAETable_f <- rbind(MAETable_f,stats$MAEci);
if (is.null(fmeth_0))
{
rcvFilter_kendall <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = univariate_correlation,featureSelection.control = list(method = "kendall",limit = 0.35,thr = 0.975),...);
}
else
{
rcvFilter_kendall <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = fmeth_0$rcvFilter_kendall$selectedFeaturesSet,...);
}
stats <- predictionStats_regression(rcvFilter_kendall$medianTest);
CorTable_f <- rbind(CorTable_f,stats$corci);
BiasTable_f <- rbind(BiasTable_f,stats$biasci);
RMSETable_f <- rbind(RMSETable_f,stats$RMSEci);
CorSpearman_f <- rbind(CorSpearman_f,stats$spearmanci);
MAETable_f <- rbind(MAETable_f,stats$MAEci);
if (is.null(fmeth_0))
{
rcvFilter_mRMR <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = mRMR.classic_FRESA,...);
}
else
{
rcvFilter_mRMR <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = fmeth_0$rcvFilter_mRMR$selectedFeaturesSet,...);
}
stats <- predictionStats_regression(rcvFilter_mRMR$medianTest);
CorTable_f <- rbind(CorTable_f,stats$corci);
BiasTable_f <- rbind(BiasTable_f,stats$biasci);
RMSETable_f <- rbind(RMSETable_f,stats$RMSEci);
CorSpearman_f <- rbind(CorSpearman_f,stats$spearmanci);
MAETable_f <- rbind(MAETable_f,stats$MAEci);
result <- list(RMSETable_filter = RMSETable_f,
CorTable_filter = CorTable_f,
BiasTable_filter = BiasTable_f,
MAE_filter = MAETable_f,
Spearam_filter = CorSpearman_f,
rcvFilter_Reference = rcvFilter_Reference,
rcvFilter_LASSO = rcvFilter_LASSO,
rcvFilter_RPART = rcvFilter_RPART,
rcvFilter_RF = rcvFilter_RF,
rcvFilter_FT = rcvFilter_FT,
rcvFilter_Wt = rcvFilter_Wt,
rcvFilter_pearson = rcvFilter_pearson,
rcvFilter_kendall = rcvFilter_kendall,
rcvFilter_mRMR = rcvFilter_mRMR
)
return(result);
}
######################Regression Algorithms####################################
if (is.null(referenceCV))
{
referenceCV <- randomCV(theData,theOutcome,BSWiMS.model,trainFraction = trainFraction,repetitions = reps,featureSelectionFunction = "Self");
referenceFilterName = "BSWiMS";
referenceName = "BSWiMS";
}
else
{
reps <- referenceCV$repetitions;
}
stats <- predictionStats_regression(referenceCV$medianTest,referenceName);
CorTable <- rbind(CorTable,stats$corci);
BiasTable <- rbind(BiasTable,stats$biasci);
RMSETable <- rbind(RMSETable,stats$RMSEci);
CorSpearman <- rbind(CorSpearman,stats$spearmanci);
MAETable <- rbind(MAETable,stats$MAEci);
rcvRF <- randomCV(theData,theOutcome,randomForest::randomForest,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = "Self");
stats <- predictionStats_regression(rcvRF$medianTest,"RF");
CorTable <- rbind(CorTable,stats$corci);
BiasTable <- rbind(BiasTable,stats$biasci);
RMSETable <- rbind(RMSETable,stats$RMSEci);
CorSpearman <- rbind(CorSpearman,stats$spearmanci);
MAETable <- rbind(MAETable,stats$MAEci);
rcvRPART <- randomCV(theData,theOutcome,rpart::rpart,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = "Self");
stats <- predictionStats_regression(rcvRPART$medianTest,"RPART");
CorTable <- rbind(CorTable,stats$corci);
BiasTable <- rbind(BiasTable,stats$biasci);
RMSETable <- rbind(RMSETable,stats$RMSEci);
CorSpearman <- rbind(CorSpearman,stats$spearmanci);
MAETable <- rbind(MAETable,stats$MAEci);
rcvLASSO <- randomCV(theData,theOutcome,LASSO_MIN,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = "Self");
stats <- predictionStats_regression(rcvLASSO$medianTest,"LASSO");
CorTable <- rbind(CorTable,stats$corci);
BiasTable <- rbind(BiasTable,stats$biasci);
RMSETable <- rbind(RMSETable,stats$RMSEci);
CorSpearman <- rbind(CorSpearman,stats$spearmanci);
MAETable <- rbind(MAETable,stats$MAEci);
rcvSVM <- randomCV(theData,theOutcome,e1071::svm,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = mRMR.classic_FRESA);
stats <- predictionStats_regression(rcvSVM$medianTest,"SVM");
CorTable <- rbind(CorTable,stats$corci);
BiasTable <- rbind(BiasTable,stats$biasci);
RMSETable <- rbind(RMSETable,stats$RMSEci);
CorSpearman <- rbind(CorSpearman,stats$spearmanci);
MAETable <- rbind(MAETable,stats$MAEci);
# Method Meta Ensemble
ens <- rowMeans(cbind(referenceCV$medianTest[,2],rcvRF$medianTest[,2],rcvLASSO$medianTest[,2],rcvSVM$medianTest[,2]))
stats <- predictionStats_regression(cbind(rcvSVM$medianTest[,1],ens),"Ensemble");
CorTable <- rbind(CorTable,stats$corci);
BiasTable <- rbind(BiasTable,stats$biasci);
RMSETable <- rbind(RMSETable,stats$RMSEci);
CorSpearman <- rbind(CorSpearman,stats$spearmanci);
MAETable <- rbind(MAETable,stats$MAEci);
######################Filters ####################################
cat("Robust Regression\n")
fmeth <- FilterMethod(MASS::rlm,"Robust Regression")
fmeth_0 <- fmeth;
CorTable_filter <- rbind(CorTable_filter,fmeth$CorTable_filter);
BiasTable_filter <- rbind(BiasTable_filter,fmeth$BiasTable_filter);
RMSETable_filter <- rbind(RMSETable_filter,fmeth$RMSETable_filter);
CorSpearman_filter <- rbind(CorSpearman_filter,fmeth$Spearam_filter);
MAETable_filter <- rbind(MAETable_filter,fmeth$MAE_filter);
cat("LASSO\n")
fmeth <- FilterMethod(LASSO_MIN,"LASSO")
CorTable_filter <- rbind(CorTable_filter,fmeth$CorTable_filter);
BiasTable_filter <- rbind(BiasTable_filter,fmeth$BiasTable_filter);
RMSETable_filter <- rbind(RMSETable_filter,fmeth$RMSETable_filter);
CorSpearman_filter <- rbind(CorSpearman_filter,fmeth$Spearam_filter);
MAETable_filter <- rbind(MAETable_filter,fmeth$MAE_filter);
cat("Filtered SVM\n")
fmeth <- FilterMethod(e1071::svm,"SVM")
CorTable_filter <- rbind(CorTable_filter,fmeth$CorTable_filter);
BiasTable_filter <- rbind(BiasTable_filter,fmeth$BiasTable_filter);
RMSETable_filter <- rbind(RMSETable_filter,fmeth$RMSETable_filter);
CorSpearman_filter <- rbind(CorSpearman_filter,fmeth$Spearam_filter);
MAETable_filter <- rbind(MAETable_filter,fmeth$MAE_filter);
cat("Random Forest\n")
fmeth <- FilterMethod(randomForest::randomForest,"Random Forest")
CorTable_filter <- rbind(CorTable_filter,fmeth$CorTable_filter);
BiasTable_filter <- rbind(BiasTable_filter,fmeth$BiasTable_filter);
RMSETable_filter <- rbind(RMSETable_filter,fmeth$RMSETable_filter);
CorSpearman_filter <- rbind(CorSpearman_filter,fmeth$Spearam_filter);
MAETable_filter <- rbind(MAETable_filter,fmeth$MAE_filter);
cat("Linear Regression\n")
fmeth <- FilterMethod(lm,"Linear Regression")
CorTable_filter <- rbind(CorTable_filter,fmeth$CorTable_filter);
BiasTable_filter <- rbind(BiasTable_filter,fmeth$BiasTable_filter);
RMSETable_filter <- rbind(RMSETable_filter,fmeth$RMSETable_filter);
CorSpearman_filter <- rbind(CorSpearman_filter,fmeth$Spearam_filter);
MAETable_filter <- rbind(MAETable_filter,fmeth$MAE_filter);
cat("Ridge Regression\n")
fmeth <- FilterMethod(LM_RIDGE_MIN,"Ridge Regression")
CorTable_filter <- rbind(CorTable_filter,fmeth$CorTable_filter);
BiasTable_filter <- rbind(BiasTable_filter,fmeth$BiasTable_filter);
RMSETable_filter <- rbind(RMSETable_filter,fmeth$RMSETable_filter);
CorSpearman_filter <- rbind(CorSpearman_filter,fmeth$Spearam_filter);
MAETable_filter <- rbind(MAETable_filter,fmeth$MAE_filter);
test_Predictions <- referenceCV$medianTest;
tnames <- rownames(test_Predictions);
testres <- rep(NA,nrow(test_Predictions));
names(testres) <- tnames;
testresnamed <- testres;
testres[tnames %in% rownames(rcvRF$medianTest)] <- rcvRF$medianTest[tnames[tnames %in% rownames(rcvRF$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres);
testres <- testresnamed;
testres[tnames %in% rownames(rcvLASSO$medianTest)] <- rcvLASSO$medianTest[tnames[tnames %in% rownames(rcvLASSO$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres);
testres <- testresnamed;
testres[tnames %in% rownames(rcvRPART$medianTest)] <- rcvRPART$medianTest[tnames[tnames %in% rownames(rcvRPART$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres);
testres <- testresnamed;
testres[tnames %in% rownames(rcvSVM$medianTest)] <- rcvSVM$medianTest[tnames[tnames %in% rownames(rcvSVM$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres);
test_Predictions <- cbind(test_Predictions,ens);
testres <- testresnamed;
testres[tnames %in% rownames(fmeth$rcvFilter_Reference$medianTest)] <- fmeth$rcvFilter_Reference$medianTest[tnames[tnames %in% rownames(fmeth$rcvFilter_Reference$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres);
testres <- testresnamed;
testres[tnames %in% rownames(fmeth$rcvFilter_LASSO$medianTest)] <- fmeth$rcvFilter_LASSO$medianTest[tnames[tnames %in% rownames(fmeth$rcvFilter_LASSO$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres);
testres <- testresnamed;
testres[tnames %in% rownames(fmeth$rcvFilter_RPART$medianTest)] <- fmeth$rcvFilter_RPART$medianTest[tnames[tnames %in% rownames(fmeth$rcvFilter_RPART$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres);
testres <- testresnamed;
testres[tnames %in% rownames(fmeth$rcvFilter_RF$medianTest)] <- fmeth$rcvFilter_RF$medianTest[tnames[tnames %in% rownames(fmeth$rcvFilter_RF$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres);
testres <- testresnamed;
testres[tnames %in% rownames(fmeth$rcvFilter_FT$medianTest)] <- fmeth$rcvFilter_FT$medianTest[tnames[tnames %in% rownames(fmeth$rcvFilter_FT$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres);
testres <- testresnamed;
testres[tnames %in% rownames(fmeth$rcvFilter_Wt$medianTest)] <- fmeth$rcvFilter_Wt$medianTest[tnames[tnames %in% rownames(fmeth$rcvFilter_Wt$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres);
testres <- testresnamed;
testres[tnames %in% rownames(fmeth$rcvFilter_pearson$medianTest)] <- fmeth$rcvFilter_pearson$medianTest[tnames[tnames %in% rownames(fmeth$rcvFilter_pearson$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres);
testres <- testresnamed;
testres[tnames %in% rownames(fmeth$rcvFilter_kendall$medianTest)] <- fmeth$rcvFilter_kendall$medianTest[tnames[tnames %in% rownames(fmeth$rcvFilter_kendall$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres);
testres <- testresnamed;
testres[tnames %in% rownames(fmeth$rcvFilter_mRMR$medianTest)] <- fmeth$rcvFilter_mRMR$medianTest[tnames[tnames %in% rownames(fmeth$rcvFilter_mRMR$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres)
colnames(test_Predictions) <- c("Outcome",referenceName,"RF","LASSO","RPART","SVM.mRMR","Ensemble",paste("RIDGE.",referenceFilterName,sep=""),"RIDGE.LASSO","RIDGE.RPART","RIDGE.RF.ref","RIDGE.FT","RIDGE.Wt","RIDGE.Pearson","RIDGE.Kendall","RIDGE.mRMR");
test_Predictions <- as.data.frame(test_Predictions)
thesets <- c("Regression Algorithm")
theMethod <- c(referenceName,"RF","RPART","LASSO","SVM","ENS")
theRegressMethod <- c("Robust Regression","LASSO","SVM","Random Forest","Linear Regression","Ridge Regression")
theFiltersets <- c(referenceFilterName,"LASSO","RPART","RF.ref","F-Test","W-Test","Pearson","Kendall","mRMR")
rownames(CorTable) <- theMethod;
rownames(BiasTable) <- theMethod;
rownames(RMSETable) <- theMethod;
rownames(CorSpearman) <- theMethod
rownames(MAETable) <- theMethod
ff <- names(referenceCV$featureFrequency)
ff <- c(ff,names(rcvLASSO$featureFrequency))
ff <- c(ff,names(rcvRPART$featureFrequency))
ff <- c(ff,names(fmeth$rcvFilter_RF$featureFrequency))
ff <- c(ff,names(fmeth$rcvFilter_FT$featureFrequency))
ff <- c(ff,names(fmeth$rcvFilter_pearson$featureFrequency))
ff <- c(ff,names(fmeth$rcvFilter_kendall$featureFrequency))
ff <- c(ff,names(fmeth$rcvFilter_mRMR$featureFrequency))
ff <- unique(ff)
Nvar <- length(ff);
selFrequency <- matrix(0,nrow = Nvar,ncol = length(theFiltersets))
# rownames(selFrequency) <- ff[1:Nvar]
rownames(selFrequency) <- ff
selnames <- rownames(selFrequency)
colnames(selFrequency) <- theFiltersets
ff <- referenceCV$featureFrequency
fnames <- selnames %in% names(ff)
selFrequency[fnames,referenceFilterName] <- ff[selnames[fnames]]
ff <- rcvLASSO$featureFrequency
fnames <- selnames %in% names(ff)
selFrequency[fnames,"LASSO"] <- ff[selnames[fnames]]
ff <- rcvRPART$featureFrequency
fnames <- selnames %in% names(ff)
selFrequency[fnames,"RPART"] <- ff[selnames[fnames]]
ff <- fmeth$rcvFilter_RF$featureFrequency
fnames <- selnames %in% names(ff)
selFrequency[fnames,"RF.ref"] <- ff[selnames[fnames]]
ff <- fmeth$rcvFilter_FT$featureFrequency
fnames <- selnames %in% names(ff)
selFrequency[fnames,"F-Test"] <- ff[selnames[fnames]]
ff <- fmeth$rcvFilter_Wt$featureFrequency
fnames <- selnames %in% names(ff)
selFrequency[fnames,"W-Test"] <- ff[selnames[fnames]]
ff <- fmeth$rcvFilter_pearson$featureFrequency
fnames <- selnames %in% names(ff)
selFrequency[fnames,"Pearson"] <- ff[selnames[fnames]]
ff <- fmeth$rcvFilter_kendall$featureFrequency
fnames <- selnames %in% names(ff)
selFrequency[fnames,"Kendall"] <- ff[selnames[fnames]]
ff <- fmeth$rcvFilter_mRMR$featureFrequency
fnames <- selnames %in% names(ff)
selFrequency[fnames,"mRMR"] <- ff[selnames[fnames]]
selFrequency <- selFrequency/reps
theMethod <- c(referenceName,"RF","RPART","LASSO","SVM","ENS")
elapcol <- names(referenceCV$theTimes) == "elapsed"
cputimes <- list(Reference = mean(referenceCV$theTimes[ elapcol ]),RF = mean(rcvRF$theTimes[ elapcol ]),RPART = mean(rcvRPART$theTimes[ elapcol ]),LASSO = mean(rcvLASSO$theTimes[ elapcol ]),SVM = mean(rcvSVM$theTimes[ elapcol ]));
jaccard_filter = list(Reference = referenceCV$jaccard,
LASSO = rcvLASSO$jaccard,
rpart = rcvRPART$jaccard,
RF.ref = fmeth$rcvFilter_RF$jaccard,
FT = fmeth$rcvFilter_FT$jaccard,
WT = fmeth$rcvFilter_Wt$jaccard,
pearson = fmeth$rcvFilter_pearson$jaccard,
kendall = fmeth$rcvFilter_kendall$jaccard,
mRMR = fmeth$rcvFilter_mRMR$jaccard
);
featsize <- unlist(lapply(jaccard_filter, `[`, c('averageLength')))
names(featsize) <- theFiltersets;
jaccard <- unlist(lapply(jaccard_filter, `[`, c('Jaccard.SM')))
names(jaccard) <- theFiltersets;
cputimes <- unlist(cputimes);
cputimes <- c(cputimes,sum(cputimes));
names(cputimes) <- theMethod;
result <- list(CorTable = CorTable,BiasTable = BiasTable,RMSETable = RMSETable,
CorTable_filter = CorTable_filter,BiasTable_filter = BiasTable_filter,RMSETable_filter = RMSETable_filter,
CorSpearman = CorSpearman, MAETable = MAETable,
CorSpearman_filter = CorSpearman_filter, MAETable_filter = MAETable_filter,
times = list(Reference = referenceCV$theTimes,RF = rcvRF$theTimes,rpart = rcvRPART$theTimes,LASSO = rcvLASSO$theTimes,SVM = rcvSVM$theTimes),
featsize = featsize,
jaccard = jaccard,
TheCVEvaluations = list(Reference = referenceCV,
RF = rcvRF,
LASSO = rcvLASSO,
RPART = rcvRPART,
SVM = rcvSVM,
RF.ref = fmeth$rcvFilter_RF,
FT = fmeth$rcvFilter_FT,
WT = fmeth$rcvFilter_Wt,
pearson = fmeth$rcvFilter_pearson,
kendall = fmeth$rcvFilter_kendall,
mRMR = fmeth$rcvFilter_mRMR
),
thesets = thesets,
theMethod = theMethod,
theRegressMethod = theRegressMethod,
theFiltersets = theFiltersets,
testPredictions = test_Predictions,
featureSelectionFrequency = selFrequency,
cpuElapsedTimes=cputimes
)
class(result) <- c("FRESA_benchmark","Regression");
return(result)
}
Try the FRESA.CAD package in your browser
Any scripts or data that you put into this service are public.
FRESA.CAD documentation built on Nov. 25, 2023, 1:07 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 RegresionBenchmark
Documented in RegresionBenchmark
RegresionBenchmark <- function(theData = NULL, theOutcome = "Class", reps = 100, trainFraction = 0.5,referenceCV = NULL,referenceName = "Reference",referenceFilterName="Reference")
{
if (!requireNamespace("e1071", quietly = TRUE)) {
install.packages("e1071", dependencies = TRUE)
}
if (!requireNamespace("randomForest", quietly = TRUE)) {
install.packages("randomForest", dependencies = TRUE)
}
if (!requireNamespace("rpart", quietly = TRUE)) {
install.packages("rpart", dependencies = TRUE)
}
if (!requireNamespace("MASS", quietly = TRUE)) {
install.packages("MASS", dependencies = TRUE)
}
if (is.null(theData))
{
if (exists("theDataSet", envir=FRESAcacheEnv))
{
theData <- get("theDataSet", envir=FRESAcacheEnv);
theOutcome <- get("theDataOutcome", envir=FRESAcacheEnv);
}
}
else
{
assign("theDataSet",theData,FRESAcacheEnv);
assign("theDataOutcome",theOutcome,FRESAcacheEnv);
}
RMSETable <- NULL
CorTable <- NULL
BiasTable <- NULL
MAETable <- NULL
CorSpearman <- NULL
RMSETable_filter <- NULL
CorTable_filter <- NULL
BiasTable_filter <- NULL
MAETable_filter <- NULL
CorSpearman_filter <- NULL
fmeth_0 <- NULL;
# par(mfrow = c(1,1));
FilterMethod <- function(regresionfun = e1071::svm, regnamefunc = "",...)
{
RMSETable_f <- NULL
CorTable_f <- NULL
MAETable_f <- NULL
CorSpearman_f<- NULL
BiasTable_f <- NULL
parm <- list(...);
rcvFilter_Reference <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = referenceCV$selectedFeaturesSet,...);
stats <- predictionStats_regression(rcvFilter_Reference$medianTest);
CorTable_f <- rbind(CorTable_f,stats$corci);
BiasTable_f <- rbind(BiasTable_f,stats$biasci);
RMSETable_f <- rbind(RMSETable_f,stats$RMSEci);
CorSpearman_f <- rbind(CorSpearman_f,stats$spearmanci);
MAETable_f <- rbind(MAETable_f,stats$MAEci);
rcvFilter_LASSO <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = rcvLASSO$selectedFeaturesSet,...);
stats <- predictionStats_regression(rcvFilter_LASSO$medianTest);
CorTable_f <- rbind(CorTable_f,stats$corci);
BiasTable_f <- rbind(BiasTable_f,stats$biasci);
RMSETable_f <- rbind(RMSETable_f,stats$RMSEci);
CorSpearman_f <- rbind(CorSpearman_f,stats$spearmanci);
MAETable_f <- rbind(MAETable_f,stats$MAEci);
rcvFilter_RPART <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = rcvRPART$selectedFeaturesSet,...);
stats <- predictionStats_regression(rcvFilter_RPART$medianTest);
CorTable_f <- rbind(CorTable_f,stats$corci);
BiasTable_f <- rbind(BiasTable_f,stats$biasci);
RMSETable_f <- rbind(RMSETable_f,stats$RMSEci);
CorSpearman_f <- rbind(CorSpearman_f,stats$spearmanci);
MAETable_f <- rbind(MAETable_f,stats$MAEci);
selectedFeaturesSet <- rcvRF$selectedFeaturesSet
for (i in 1:length(selectedFeaturesSet))
{
if (length(referenceCV$selectedFeaturesSet[[i]]) > 1)
{
if (length(selectedFeaturesSet[[i]]) > length(referenceCV$selectedFeaturesSet[[i]]))
{
selectedFeaturesSet[[i]] <- selectedFeaturesSet[[i]][1:length(referenceCV$selectedFeaturesSet[[i]])];
}
}
else # the top five or RF
{
warning ("Less than 2 features, then will keep the top five of RF\n")
if (length(selectedFeaturesSet[[i]]) > 5)
{
selectedFeaturesSet[[i]] <- selectedFeaturesSet[[i]][1:5];
}
}
}
rcvFilter_RF <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = selectedFeaturesSet,...);
stats <- predictionStats_regression(rcvFilter_RF$medianTest);
CorTable_f <- rbind(CorTable_f,stats$corci);
BiasTable_f <- rbind(BiasTable_f,stats$biasci);
RMSETable_f <- rbind(RMSETable_f,stats$RMSEci);
CorSpearman_f <- rbind(CorSpearman_f,stats$spearmanci);
MAETable_f <- rbind(MAETable_f,stats$MAEci);
if (is.null(fmeth_0))
{
rcvFilter_FT <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = univariate_residual,featureSelection.control = list(uniTest = "Ftest",limit = 0.35,thr = 0.975),...);
}
else
{
rcvFilter_FT <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = fmeth_0$rcvFilter_FT$selectedFeaturesSet,...);
}
stats <- predictionStats_regression(rcvFilter_FT$medianTest);
CorTable_f <- rbind(CorTable_f,stats$corci);
BiasTable_f <- rbind(BiasTable_f,stats$biasci);
RMSETable_f <- rbind(RMSETable_f,stats$RMSEci);
CorSpearman_f <- rbind(CorSpearman_f,stats$spearmanci);
MAETable_f <- rbind(MAETable_f,stats$MAEci);
if (is.null(fmeth_0))
{
rcvFilter_Wt <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = univariate_residual,featureSelection.control = list(uniTest = "Wilcox",limit = 0.35,thr = 0.975),...);
}
else
{
rcvFilter_Wt <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = fmeth_0$rcvFilter_Wt$selectedFeaturesSet,...);
}
stats <- predictionStats_regression(rcvFilter_Wt$medianTest);
CorTable_f <- rbind(CorTable_f,stats$corci);
BiasTable_f <- rbind(BiasTable_f,stats$biasci);
RMSETable_f <- rbind(RMSETable_f,stats$RMSEci);
CorSpearman_f <- rbind(CorSpearman_f,stats$spearmanci);
MAETable_f <- rbind(MAETable_f,stats$MAEci);
if (is.null(fmeth_0))
{
rcvFilter_pearson <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = univariate_correlation,featureSelection.control = list(method = "pearson",limit = 0.35,thr = 0.975),...);
}
else
{
rcvFilter_pearson <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = fmeth_0$rcvFilter_pearson$selectedFeaturesSet,...);
}
stats <- predictionStats_regression(rcvFilter_pearson$medianTest);
CorTable_f <- rbind(CorTable_f,stats$corci);
BiasTable_f <- rbind(BiasTable_f,stats$biasci);
RMSETable_f <- rbind(RMSETable_f,stats$RMSEci);
CorSpearman_f <- rbind(CorSpearman_f,stats$spearmanci);
MAETable_f <- rbind(MAETable_f,stats$MAEci);
if (is.null(fmeth_0))
{
rcvFilter_kendall <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = univariate_correlation,featureSelection.control = list(method = "kendall",limit = 0.35,thr = 0.975),...);
}
else
{
rcvFilter_kendall <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = fmeth_0$rcvFilter_kendall$selectedFeaturesSet,...);
}
stats <- predictionStats_regression(rcvFilter_kendall$medianTest);
CorTable_f <- rbind(CorTable_f,stats$corci);
BiasTable_f <- rbind(BiasTable_f,stats$biasci);
RMSETable_f <- rbind(RMSETable_f,stats$RMSEci);
CorSpearman_f <- rbind(CorSpearman_f,stats$spearmanci);
MAETable_f <- rbind(MAETable_f,stats$MAEci);
if (is.null(fmeth_0))
{
rcvFilter_mRMR <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = mRMR.classic_FRESA,...);
}
else
{
rcvFilter_mRMR <- randomCV(theData,theOutcome,regresionfun,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = fmeth_0$rcvFilter_mRMR$selectedFeaturesSet,...);
}
stats <- predictionStats_regression(rcvFilter_mRMR$medianTest);
CorTable_f <- rbind(CorTable_f,stats$corci);
BiasTable_f <- rbind(BiasTable_f,stats$biasci);
RMSETable_f <- rbind(RMSETable_f,stats$RMSEci);
CorSpearman_f <- rbind(CorSpearman_f,stats$spearmanci);
MAETable_f <- rbind(MAETable_f,stats$MAEci);
result <- list(RMSETable_filter = RMSETable_f,
CorTable_filter = CorTable_f,
BiasTable_filter = BiasTable_f,
MAE_filter = MAETable_f,
Spearam_filter = CorSpearman_f,
rcvFilter_Reference = rcvFilter_Reference,
rcvFilter_LASSO = rcvFilter_LASSO,
rcvFilter_RPART = rcvFilter_RPART,
rcvFilter_RF = rcvFilter_RF,
rcvFilter_FT = rcvFilter_FT,
rcvFilter_Wt = rcvFilter_Wt,
rcvFilter_pearson = rcvFilter_pearson,
rcvFilter_kendall = rcvFilter_kendall,
rcvFilter_mRMR = rcvFilter_mRMR
)
return(result);
}
######################Regression Algorithms####################################
if (is.null(referenceCV))
{
referenceCV <- randomCV(theData,theOutcome,BSWiMS.model,trainFraction = trainFraction,repetitions = reps,featureSelectionFunction = "Self");
referenceFilterName = "BSWiMS";
referenceName = "BSWiMS";
}
else
{
reps <- referenceCV$repetitions;
}
stats <- predictionStats_regression(referenceCV$medianTest,referenceName);
CorTable <- rbind(CorTable,stats$corci);
BiasTable <- rbind(BiasTable,stats$biasci);
RMSETable <- rbind(RMSETable,stats$RMSEci);
CorSpearman <- rbind(CorSpearman,stats$spearmanci);
MAETable <- rbind(MAETable,stats$MAEci);
rcvRF <- randomCV(theData,theOutcome,randomForest::randomForest,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = "Self");
stats <- predictionStats_regression(rcvRF$medianTest,"RF");
CorTable <- rbind(CorTable,stats$corci);
BiasTable <- rbind(BiasTable,stats$biasci);
RMSETable <- rbind(RMSETable,stats$RMSEci);
CorSpearman <- rbind(CorSpearman,stats$spearmanci);
MAETable <- rbind(MAETable,stats$MAEci);
rcvRPART <- randomCV(theData,theOutcome,rpart::rpart,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = "Self");
stats <- predictionStats_regression(rcvRPART$medianTest,"RPART");
CorTable <- rbind(CorTable,stats$corci);
BiasTable <- rbind(BiasTable,stats$biasci);
RMSETable <- rbind(RMSETable,stats$RMSEci);
CorSpearman <- rbind(CorSpearman,stats$spearmanci);
MAETable <- rbind(MAETable,stats$MAEci);
rcvLASSO <- randomCV(theData,theOutcome,LASSO_MIN,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = "Self");
stats <- predictionStats_regression(rcvLASSO$medianTest,"LASSO");
CorTable <- rbind(CorTable,stats$corci);
BiasTable <- rbind(BiasTable,stats$biasci);
RMSETable <- rbind(RMSETable,stats$RMSEci);
CorSpearman <- rbind(CorSpearman,stats$spearmanci);
MAETable <- rbind(MAETable,stats$MAEci);
rcvSVM <- randomCV(theData,theOutcome,e1071::svm,trainSampleSets = referenceCV$trainSamplesSets,featureSelectionFunction = mRMR.classic_FRESA);
stats <- predictionStats_regression(rcvSVM$medianTest,"SVM");
CorTable <- rbind(CorTable,stats$corci);
BiasTable <- rbind(BiasTable,stats$biasci);
RMSETable <- rbind(RMSETable,stats$RMSEci);
CorSpearman <- rbind(CorSpearman,stats$spearmanci);
MAETable <- rbind(MAETable,stats$MAEci);
# Method Meta Ensemble
ens <- rowMeans(cbind(referenceCV$medianTest[,2],rcvRF$medianTest[,2],rcvLASSO$medianTest[,2],rcvSVM$medianTest[,2]))
stats <- predictionStats_regression(cbind(rcvSVM$medianTest[,1],ens),"Ensemble");
CorTable <- rbind(CorTable,stats$corci);
BiasTable <- rbind(BiasTable,stats$biasci);
RMSETable <- rbind(RMSETable,stats$RMSEci);
CorSpearman <- rbind(CorSpearman,stats$spearmanci);
MAETable <- rbind(MAETable,stats$MAEci);
######################Filters ####################################
cat("Robust Regression\n")
fmeth <- FilterMethod(MASS::rlm,"Robust Regression")
fmeth_0 <- fmeth;
CorTable_filter <- rbind(CorTable_filter,fmeth$CorTable_filter);
BiasTable_filter <- rbind(BiasTable_filter,fmeth$BiasTable_filter);
RMSETable_filter <- rbind(RMSETable_filter,fmeth$RMSETable_filter);
CorSpearman_filter <- rbind(CorSpearman_filter,fmeth$Spearam_filter);
MAETable_filter <- rbind(MAETable_filter,fmeth$MAE_filter);
cat("LASSO\n")
fmeth <- FilterMethod(LASSO_MIN,"LASSO")
CorTable_filter <- rbind(CorTable_filter,fmeth$CorTable_filter);
BiasTable_filter <- rbind(BiasTable_filter,fmeth$BiasTable_filter);
RMSETable_filter <- rbind(RMSETable_filter,fmeth$RMSETable_filter);
CorSpearman_filter <- rbind(CorSpearman_filter,fmeth$Spearam_filter);
MAETable_filter <- rbind(MAETable_filter,fmeth$MAE_filter);
cat("Filtered SVM\n")
fmeth <- FilterMethod(e1071::svm,"SVM")
CorTable_filter <- rbind(CorTable_filter,fmeth$CorTable_filter);
BiasTable_filter <- rbind(BiasTable_filter,fmeth$BiasTable_filter);
RMSETable_filter <- rbind(RMSETable_filter,fmeth$RMSETable_filter);
CorSpearman_filter <- rbind(CorSpearman_filter,fmeth$Spearam_filter);
MAETable_filter <- rbind(MAETable_filter,fmeth$MAE_filter);
cat("Random Forest\n")
fmeth <- FilterMethod(randomForest::randomForest,"Random Forest")
CorTable_filter <- rbind(CorTable_filter,fmeth$CorTable_filter);
BiasTable_filter <- rbind(BiasTable_filter,fmeth$BiasTable_filter);
RMSETable_filter <- rbind(RMSETable_filter,fmeth$RMSETable_filter);
CorSpearman_filter <- rbind(CorSpearman_filter,fmeth$Spearam_filter);
MAETable_filter <- rbind(MAETable_filter,fmeth$MAE_filter);
cat("Linear Regression\n")
fmeth <- FilterMethod(lm,"Linear Regression")
CorTable_filter <- rbind(CorTable_filter,fmeth$CorTable_filter);
BiasTable_filter <- rbind(BiasTable_filter,fmeth$BiasTable_filter);
RMSETable_filter <- rbind(RMSETable_filter,fmeth$RMSETable_filter);
CorSpearman_filter <- rbind(CorSpearman_filter,fmeth$Spearam_filter);
MAETable_filter <- rbind(MAETable_filter,fmeth$MAE_filter);
cat("Ridge Regression\n")
fmeth <- FilterMethod(LM_RIDGE_MIN,"Ridge Regression")
CorTable_filter <- rbind(CorTable_filter,fmeth$CorTable_filter);
BiasTable_filter <- rbind(BiasTable_filter,fmeth$BiasTable_filter);
RMSETable_filter <- rbind(RMSETable_filter,fmeth$RMSETable_filter);
CorSpearman_filter <- rbind(CorSpearman_filter,fmeth$Spearam_filter);
MAETable_filter <- rbind(MAETable_filter,fmeth$MAE_filter);
test_Predictions <- referenceCV$medianTest;
tnames <- rownames(test_Predictions);
testres <- rep(NA,nrow(test_Predictions));
names(testres) <- tnames;
testresnamed <- testres;
testres[tnames %in% rownames(rcvRF$medianTest)] <- rcvRF$medianTest[tnames[tnames %in% rownames(rcvRF$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres);
testres <- testresnamed;
testres[tnames %in% rownames(rcvLASSO$medianTest)] <- rcvLASSO$medianTest[tnames[tnames %in% rownames(rcvLASSO$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres);
testres <- testresnamed;
testres[tnames %in% rownames(rcvRPART$medianTest)] <- rcvRPART$medianTest[tnames[tnames %in% rownames(rcvRPART$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres);
testres <- testresnamed;
testres[tnames %in% rownames(rcvSVM$medianTest)] <- rcvSVM$medianTest[tnames[tnames %in% rownames(rcvSVM$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres);
test_Predictions <- cbind(test_Predictions,ens);
testres <- testresnamed;
testres[tnames %in% rownames(fmeth$rcvFilter_Reference$medianTest)] <- fmeth$rcvFilter_Reference$medianTest[tnames[tnames %in% rownames(fmeth$rcvFilter_Reference$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres);
testres <- testresnamed;
testres[tnames %in% rownames(fmeth$rcvFilter_LASSO$medianTest)] <- fmeth$rcvFilter_LASSO$medianTest[tnames[tnames %in% rownames(fmeth$rcvFilter_LASSO$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres);
testres <- testresnamed;
testres[tnames %in% rownames(fmeth$rcvFilter_RPART$medianTest)] <- fmeth$rcvFilter_RPART$medianTest[tnames[tnames %in% rownames(fmeth$rcvFilter_RPART$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres);
testres <- testresnamed;
testres[tnames %in% rownames(fmeth$rcvFilter_RF$medianTest)] <- fmeth$rcvFilter_RF$medianTest[tnames[tnames %in% rownames(fmeth$rcvFilter_RF$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres);
testres <- testresnamed;
testres[tnames %in% rownames(fmeth$rcvFilter_FT$medianTest)] <- fmeth$rcvFilter_FT$medianTest[tnames[tnames %in% rownames(fmeth$rcvFilter_FT$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres);
testres <- testresnamed;
testres[tnames %in% rownames(fmeth$rcvFilter_Wt$medianTest)] <- fmeth$rcvFilter_Wt$medianTest[tnames[tnames %in% rownames(fmeth$rcvFilter_Wt$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres);
testres <- testresnamed;
testres[tnames %in% rownames(fmeth$rcvFilter_pearson$medianTest)] <- fmeth$rcvFilter_pearson$medianTest[tnames[tnames %in% rownames(fmeth$rcvFilter_pearson$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres);
testres <- testresnamed;
testres[tnames %in% rownames(fmeth$rcvFilter_kendall$medianTest)] <- fmeth$rcvFilter_kendall$medianTest[tnames[tnames %in% rownames(fmeth$rcvFilter_kendall$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres);
testres <- testresnamed;
testres[tnames %in% rownames(fmeth$rcvFilter_mRMR$medianTest)] <- fmeth$rcvFilter_mRMR$medianTest[tnames[tnames %in% rownames(fmeth$rcvFilter_mRMR$medianTest)],2];
test_Predictions <- cbind(test_Predictions,testres)
colnames(test_Predictions) <- c("Outcome",referenceName,"RF","LASSO","RPART","SVM.mRMR","Ensemble",paste("RIDGE.",referenceFilterName,sep=""),"RIDGE.LASSO","RIDGE.RPART","RIDGE.RF.ref","RIDGE.FT","RIDGE.Wt","RIDGE.Pearson","RIDGE.Kendall","RIDGE.mRMR");
test_Predictions <- as.data.frame(test_Predictions)
thesets <- c("Regression Algorithm")
theMethod <- c(referenceName,"RF","RPART","LASSO","SVM","ENS")
theRegressMethod <- c("Robust Regression","LASSO","SVM","Random Forest","Linear Regression","Ridge Regression")
theFiltersets <- c(referenceFilterName,"LASSO","RPART","RF.ref","F-Test","W-Test","Pearson","Kendall","mRMR")
rownames(CorTable) <- theMethod;
rownames(BiasTable) <- theMethod;
rownames(RMSETable) <- theMethod;
rownames(CorSpearman) <- theMethod
rownames(MAETable) <- theMethod
ff <- names(referenceCV$featureFrequency)
ff <- c(ff,names(rcvLASSO$featureFrequency))
ff <- c(ff,names(rcvRPART$featureFrequency))
ff <- c(ff,names(fmeth$rcvFilter_RF$featureFrequency))
ff <- c(ff,names(fmeth$rcvFilter_FT$featureFrequency))
ff <- c(ff,names(fmeth$rcvFilter_pearson$featureFrequency))
ff <- c(ff,names(fmeth$rcvFilter_kendall$featureFrequency))
ff <- c(ff,names(fmeth$rcvFilter_mRMR$featureFrequency))
ff <- unique(ff)
Nvar <- length(ff);
selFrequency <- matrix(0,nrow = Nvar,ncol = length(theFiltersets))
# rownames(selFrequency) <- ff[1:Nvar]
rownames(selFrequency) <- ff
selnames <- rownames(selFrequency)
colnames(selFrequency) <- theFiltersets
ff <- referenceCV$featureFrequency
fnames <- selnames %in% names(ff)
selFrequency[fnames,referenceFilterName] <- ff[selnames[fnames]]
ff <- rcvLASSO$featureFrequency
fnames <- selnames %in% names(ff)
selFrequency[fnames,"LASSO"] <- ff[selnames[fnames]]
ff <- rcvRPART$featureFrequency
fnames <- selnames %in% names(ff)
selFrequency[fnames,"RPART"] <- ff[selnames[fnames]]
ff <- fmeth$rcvFilter_RF$featureFrequency
fnames <- selnames %in% names(ff)
selFrequency[fnames,"RF.ref"] <- ff[selnames[fnames]]
ff <- fmeth$rcvFilter_FT$featureFrequency
fnames <- selnames %in% names(ff)
selFrequency[fnames,"F-Test"] <- ff[selnames[fnames]]
ff <- fmeth$rcvFilter_Wt$featureFrequency
fnames <- selnames %in% names(ff)
selFrequency[fnames,"W-Test"] <- ff[selnames[fnames]]
ff <- fmeth$rcvFilter_pearson$featureFrequency
fnames <- selnames %in% names(ff)
selFrequency[fnames,"Pearson"] <- ff[selnames[fnames]]
ff <- fmeth$rcvFilter_kendall$featureFrequency
fnames <- selnames %in% names(ff)
selFrequency[fnames,"Kendall"] <- ff[selnames[fnames]]
ff <- fmeth$rcvFilter_mRMR$featureFrequency
fnames <- selnames %in% names(ff)
selFrequency[fnames,"mRMR"] <- ff[selnames[fnames]]
selFrequency <- selFrequency/reps
theMethod <- c(referenceName,"RF","RPART","LASSO","SVM","ENS")
elapcol <- names(referenceCV$theTimes) == "elapsed"
cputimes <- list(Reference = mean(referenceCV$theTimes[ elapcol ]),RF = mean(rcvRF$theTimes[ elapcol ]),RPART = mean(rcvRPART$theTimes[ elapcol ]),LASSO = mean(rcvLASSO$theTimes[ elapcol ]),SVM = mean(rcvSVM$theTimes[ elapcol ]));
jaccard_filter = list(Reference = referenceCV$jaccard,
LASSO = rcvLASSO$jaccard,
rpart = rcvRPART$jaccard,
RF.ref = fmeth$rcvFilter_RF$jaccard,
FT = fmeth$rcvFilter_FT$jaccard,
WT = fmeth$rcvFilter_Wt$jaccard,
pearson = fmeth$rcvFilter_pearson$jaccard,
kendall = fmeth$rcvFilter_kendall$jaccard,
mRMR = fmeth$rcvFilter_mRMR$jaccard
);
featsize <- unlist(lapply(jaccard_filter, `[`, c('averageLength')))
names(featsize) <- theFiltersets;
jaccard <- unlist(lapply(jaccard_filter, `[`, c('Jaccard.SM')))
names(jaccard) <- theFiltersets;
cputimes <- unlist(cputimes);
cputimes <- c(cputimes,sum(cputimes));
names(cputimes) <- theMethod;
result <- list(CorTable = CorTable,BiasTable = BiasTable,RMSETable = RMSETable,
CorTable_filter = CorTable_filter,BiasTable_filter = BiasTable_filter,RMSETable_filter = RMSETable_filter,
CorSpearman = CorSpearman, MAETable = MAETable,
CorSpearman_filter = CorSpearman_filter, MAETable_filter = MAETable_filter,
times = list(Reference = referenceCV$theTimes,RF = rcvRF$theTimes,rpart = rcvRPART$theTimes,LASSO = rcvLASSO$theTimes,SVM = rcvSVM$theTimes),
featsize = featsize,
jaccard = jaccard,
TheCVEvaluations = list(Reference = referenceCV,
RF = rcvRF,
LASSO = rcvLASSO,
RPART = rcvRPART,
SVM = rcvSVM,
RF.ref = fmeth$rcvFilter_RF,
FT = fmeth$rcvFilter_FT,
WT = fmeth$rcvFilter_Wt,
pearson = fmeth$rcvFilter_pearson,
kendall = fmeth$rcvFilter_kendall,
mRMR = fmeth$rcvFilter_mRMR
),
thesets = thesets,
theMethod = theMethod,
theRegressMethod = theRegressMethod,
theFiltersets = theFiltersets,
testPredictions = test_Predictions,
featureSelectionFrequency = selFrequency,
cpuElapsedTimes=cputimes
)
class(result) <- c("FRESA_benchmark","Regression");
return(result)
}
Try the FRESA.CAD package in your browser
Any scripts or data that you put into this service are public.
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.