Better use of preload data

R/commonMachineLearningClassification.R

@@ -42,12 +42,7 @@
 }
 
 .mlClassificationReadData <- function(dataset, options) {
-  if (is.null(dataset)) {
-    dataset <- .readDataClassificationRegressionAnalyses(dataset, options)
-  }
-  if (length(unlist(options[["predictors"]])) > 0 && options[["scaleVariables"]]) {
-    dataset[, options[["predictors"]]] <- .scaleNumericData(dataset[, options[["predictors"]], drop = FALSE])
-  }
+  dataset <- .readDataClassificationRegressionAnalyses(dataset, options, include_weights = FALSE)
   if (options[["target"]] != "") {
     dataset[, options[["target"]]] <- factor(dataset[, options[["target"]]], ordered = FALSE)
   }

R/commonMachineLearningClustering.R

@@ -33,9 +33,7 @@
 .mlClusteringReadData <- function(dataset, options) {
   predictors <- unlist(options[["predictors"]])
   predictors <- predictors[predictors != ""]
-  if (is.null(dataset)) {
-    dataset <- .readAndAddCompleteRowIndices(options, "predictors")
-  }
+  dataset <- jaspBase::excludeNaListwise(dataset, predictors)
   if (options[["scaleVariables"]] && length(unlist(options[["predictors"]])) > 0) {
     dataset <- .scaleNumericData(dataset)
   }

R/commonMachineLearningRegression.R

@@ -44,23 +44,39 @@
   return(opt)
 }
 
-.readDataRegressionAnalyses <- function(dataset, options, jaspResults) {
-  if (is.null(dataset)) {
-    dataset <- .readDataClassificationRegressionAnalyses(dataset, options)
-  }
-  if (length(unlist(options[["predictors"]])) > 0 && options[["scaleVariables"]]) {
-    dataset[, options[["predictors"]]] <- .scaleNumericData(dataset[, options[["predictors"]], drop = FALSE])
-  }
+.readDataRegressionAnalyses <- function(dataset, options, jaspResults, include_weights = FALSE) {
+  dataset <- .readDataClassificationRegressionAnalyses(dataset, options, include_weights)
   return(dataset)
 }
 
-.readDataClassificationRegressionAnalyses <- function(dataset, options) {
+.readDataClassificationRegressionAnalyses <- function(dataset, options, include_weights) {
 
+  target <- NULL
+  weights <- NULL
   testSetIndicator <- NULL
-  if (options[["testSetIndicatorVariable"]] != "" && options[["holdoutData"]] == "testSetIndicator")
-    testSetIndicator <- "testSetIndicatorVariable"
+  if (options[["target"]] != "") {
+    target <- options[["target"]]
+  }
+  if (include_weights && options[["weights"]] != "") {
+    weights <- options[["weights"]]
+  }
+  if (options[["testSetIndicatorVariable"]] != "" && options[["holdoutData"]] == "testSetIndicator") {
+    testSetIndicator <- options[["testSetIndicatorVariable"]]
+  }
+
+  predictors <- unlist(options["predictors"])
+  predictors <- predictors[predictors != ""]
+  dataset <- jaspBase::excludeNaListwise(dataset, c(target, predictors, weights, testSetIndicator))
 
-  return(.readAndAddCompleteRowIndices(options, c("target", "predictors"), testSetIndicator))
+  # Scale numeric predictors
+  if (length(unlist(options[["predictors"]])) > 0 && options[["scaleVariables"]]) {
+    dataset[, options[["predictors"]]] <- .scaleNumericData(dataset[, options[["predictors"]], drop = FALSE])
+  }
+  # Make sure the test set indicator is numeric
+  if (options[["testSetIndicatorVariable"]] != "" && options[["holdoutData"]] == "testSetIndicator")
+    dataset[[options[["testSetIndicatorVariable"]]]] <- as.numeric(dataset[[options[["testSetIndicatorVariable"]]]])
+
+  return(dataset)
 }
 
 .readAndAddCompleteRowIndices <- function(options, optionNames = NULL, optionNamesAsNumeric = NULL) {
@@ -72,9 +88,7 @@
         options[[name2]] <- rep("scale", length(options[[name]]))
     }
   dataset <- jaspBase::readDataSetByVariableTypes(options, c(optionNames, optionNamesAsNumeric))
-  complete.index <- which(complete.cases(dataset))
-  dataset <- na.omit(dataset)
-  rownames(dataset) <- as.character(complete.index)
+  dataset <- jaspBase::excludeNaListwise(dataset, c(options[["target"]], options[["predictors"]]))
   return(dataset)
 }
 

R/mlPrediction.R

@@ -422,7 +422,7 @@ is.jaspMachineLearning <- function(x) {
   selection <- predictions[indexes]
   cols <- list(row = indexes, pred = selection)
   if (options[["predictionsTableFeatures"]]) {
-    for (i in model[["jaspVars"]][["encoded"]]$predictors) {
+    for (i in colnames(dataset)) {
       if (.columnIsNominal(i)) {
         table$addColumnInfo(name = i, title = i, type = "string")
         var <- levels(dataset[[i]])[dataset[[i]]]

R/mlRegressionLinear.R

@@ -18,7 +18,7 @@
 mlRegressionLinear <- function(jaspResults, dataset, options, ...) {
 
   # Preparatory work
-  dataset <- .mlRegressionRegularizedReadData(dataset, options)
+  dataset <- .readDataRegressionAnalyses(dataset, options, include_weights = TRUE)
   .mlRegressionErrorHandling(dataset, options, type = "lm")
 
   # Check if analysis is ready to run

R/mlRegressionRegularized.R

@@ -18,7 +18,7 @@
 mlRegressionRegularized <- function(jaspResults, dataset, options, ...) {
 
   # Preparatory work
-  dataset <- .mlRegressionRegularizedReadData(dataset, options)
+  dataset <- .readDataRegressionAnalyses(dataset, options, include_weights = TRUE)
   .mlRegressionErrorHandling(dataset, options, type = "regularized")
 
   # Check if analysis is ready to run
@@ -58,31 +58,6 @@ mlRegressionRegularized <- function(jaspResults, dataset, options, ...) {
   .mlRegressionRegularizedPlotLambda(options, jaspResults, ready, position = 10)
 }
 
-# Read dataset
-.mlRegressionRegularizedReadData <- function(dataset, options) {
-  target <- NULL
-  weights <- NULL
-  testSetIndicator <- NULL
-  if (options[["target"]] != "") {
-    target <- options[["target"]]
-  }
-  if (options[["weights"]] != "") {
-    weights <- options[["weights"]]
-  }
-  if (options[["testSetIndicatorVariable"]] != "" && options[["holdoutData"]] == "testSetIndicator")
-    testSetIndicator <- "testSetIndicatorVariable"
-
-  predictors <- unlist(options["predictors"])
-  predictors <- predictors[predictors != ""]
-  if (is.null(dataset)) {
-    dataset <- .readAndAddCompleteRowIndices(options, c("target", "predictors", "weights"), testSetIndicator)
-  }
-  if (length(unlist(options[["predictors"]])) > 0 && options[["scaleVariables"]]) {
-    dataset[, options[["predictors"]]] <- .scaleNumericData(dataset[, options[["predictors"]], drop = FALSE])
-  }
-  return(dataset)
-}
-
 .regularizedRegression <- function(dataset, options, jaspResults) {
   # Set model-specific parameters
   alpha <- switch(options[["penalty"]],

inst/help/mlPrediction.md

@@ -1,6 +1,8 @@
 Prediction
 ===
 
+The prediction analysis enables you to load a trained machine learning model and apply it to new data. It is important that the features in the new dataset have the same names as in the original dataset used for training.
+
 ### Input
 
 #### Trained Model

inst/help/mlPrediction_nl.md

@@ -1,6 +1,8 @@
 Voorspellen
 ===
 
+Met de voorspellingsanalyse kun je een getraind machine-learningmodel laden en toepassen op nieuwe gegevens. Het is belangrijk dat de kenmerken in de nieuwe dataset dezelfde namen hebben als in de oorspronkelijke dataset die voor de training is gebruikt.
+
 ### Invoer
 
 #### Getraind model

inst/qml/mlPrediction.qml

@@ -26,7 +26,7 @@ import "./common/tables" as TAB
 
 Form 
 {
-	info: qsTr("The prediction analysis enables you to load a trained machine learning model and apply it to new data.")
+	info: qsTr("The prediction analysis enables you to load a trained machine learning model and apply it to new data. It is important that the features in the new dataset have the same names as in the original dataset used for training.")
 
 	FileSelector
 	{
@@ -53,7 +53,7 @@ Form
 			id:									predictors
 			name:								"predictors"
 			title:								qsTr("Features")
-			allowedColumns:						["scale", "ordinal", "nominal"]
+			allowedColumns:						["scale", "nominal"]
 			allowAnalysisOwnComputedColumns:	false
 		}
 	}