CrossValidateModels: Cross validates models to reproduce results in journal and...

In sevvandi/outselect: Algorithm selection for unsupervised outlier detection

Description

Cross validates models to reproduce results in journal and conference papers.

Usage

CrossValidateModels(d, p, rocpr = 1, s = 1, n = 5)

Arguments

d	If d=1 then we take the Min_Max performance values, if d=2 then performance values from all normalization methods are considered. Input values for d are only 1,2.
p	If p=1 then we take binary values based on absolute performance, i.e. if performance > 0.8, if p=2 the relative binary performance values are used. Input values for p are only 1,2.
rocpr	For all normalization methods, if rocpr=1 then area under ROC curve is used, if rocpr=2 then area under PR curve is used. For Min_Max method, i.e. if d=1, then only ROC values are used.
s	If s=1 then we train the models on a preferred subset of features. If s=2 the models are trained on all features, which takes considerably longer. Default value s=1. Input values for s are only 1,2.
n	The number of folds in cross validation.

Value

A list containing the following results from the cross validated models:

def_acc

The default accuracy we get if we predict the method is not good for all instances. This is the percentage of the majority class.

results

The n-fold cross valdation results.

mean_acc

The mean n-fold cross valdation results.

Examples

## Not run: 
out <- CrossValidateModels(1,1,1,1,5)
out$mean_acc

## End(Not run)

sevvandi/outselect documentation built on June 1, 2019, 3:58 a.m.