getMetricsPCA: Performance Measures with PCA pre-processing

In khadijaaziz/icardaFIGSr: Subsetting using Focused Identification of the Germplasm Strategy (FIGS)

Description

getMetricsPCA allows to obtain performance measures from Confusion Matrix for algorithms with PCA pre-processing,it returns a data frame containing performance measures from the confusion matrix given by the caret package when algorithms have been run with PCA pre-processing.

Usage

getMetricsPCA(yhat, y, classtype, model)

Arguments

yhat	expression. The vector of predicted values.
y	expression. The class variable.
classtype	character or numeric. The number of levels in y.
model	expression. The model object to which output of the model has been assigned.

Details

Works with target variables that have two, three, four, six or eight classes. Similar to getMetrics but used in the case where models have been run with PCA specified as an option for the preProcess argument in the train function of caret.

Value

Outputs an object with performance measures calculated from the confusion matrix given by the caret package. A data frame is the resulting output with the first column giving the name of the performance measure, and the second column giving the corresponding value.

Author(s)

Khadija Aziz, Zainab Azough, Zakaria Kehel, Bancy Ngatia

See Also

confusionMatrix, predict.train

Examples

if(interactive()){
 # Obtain predictions from several previously run models
 dataX <- subset(data, select = -y)
 pred.knn <- predict(model.knn, newdata = dataX)
 pred.rf <- predict(model.rf, newdata = dataX)

 # Get metrics for several algorithms
 metrics.knn <- getMetricsPCA(y = data$y, yhat = pred.knn,
                              classtype = 2, model = model.knn)
 metrics.rf <- getMetricsPCA(y = data$y, yhat = pred.rf,
                             classtype = 2, model = model.rf)

 # Indexing for 2-class models to remove extra column with
 # names of performance measures
 metrics.all <- cbind(metrics.knn, metrics.rf[ , 2])

 # No indexing needed for 3-, 4-, 6- or 8-class models
 metrics.all <- cbind(metrics.knn, metrics.rf)
 }

khadijaaziz/icardaFIGSr documentation built on Dec. 21, 2021, 6:38 a.m.