expGetDuplicated

Create a new experiment with only the duplicated rows

Usage

expGetDuplicated(e, tol = 1e-09)

Arguments

e
The experiment to check for duplicated rows
tol
The tolerance for numeric values to check if two outputs are numerically equal or not.

Value

A new experiment containing the duplicated rows

Description

This function computes the duplicated rows attending to the method, problem and input parameters (but not the outputs). The resulting experiment will contain these duplicated rows.

Details

If duplicated rows show different outputs the function will launch a a warning message indicating how many of them differ in the outputs from the original row, the extent to what two rows are divergent in their output can be parametrized.

This function is useful to determine the consistency of the experiment, as a measure to sanitice the original data source if needed,

Examples

# We duplicate some of the rows of a given experiment: e <- expCreate(wekaExperiment, parameters="fold", name="Test Experiment") redundant <- expCreate(wekaExperiment[wekaExperiment$method=="NaiveBayes",], parameters="fold", name="Test Experiment") e2 <- expConcat(e,redundant)
Warning message: 300 duplicated rows. 0 differ in the outputs (using a tolerance of 1.0000e-09 to compare the outputs)
# Now we check for duplicates: expGetDuplicated(e2)
Warning message: 300 duplicated rows. 0 differ in the outputs (using a tolerance of 1.0000e-09 to compare the outputs)
#Experiment name: Test Experiment+Test Experiment #method: J48, NaiveBayes, OneR, RandomForest #problem: anneal, audiology, balance-scale, car, glass, horsecolic, hypothyroid, ionosphere, liver-disorders, lymph, primary-tumor, soybean, vehicle, vote, vowel #parameters: *) featureSelection [no,yes] *) fold [0,1,2,3,4,5,6,7,8,9] #outputs: accuracy, trainingTime

Back to top

Built by staticdocs. Styled with bootstrap.