train_model: Train binary classifier

In NLeSC/EEG-epilepsy-diagnosis: Epilepsy detection with emotiv data

Description

Loads EEG data and writes cleaned data to an output directory

Usage

train_model(DATtrain,LABtrain,DATval,LABval,featuredict,classifier="rf")

Arguments

DATtrain	Training feature values as produced by extract_features and split up by split_data
LABtrain	Training labels as derived from split_data
DATval	Validation feature values as produced by extract_features and split up by split_data
LABval	Validation labels as derived from split_data
featuredict	Dataframe with all overview of all available combinations of wavelet types, wavelet levels, extracted signal features, and aggregation type.
classifier	type of classification model, defaul is "rf" for random forrest. Alternative value could be "lg" for logistic regression

Value

result	classificiation performance in the validation set for each value of the hyperparameter
best_model	classification model as produced by caret function train
fes	column indices of the DAT corresponding to the features used for the best model

Author(s)

Vincent T van Hees

Examples

## Not run: 
  data(data.eeg)
  data(data.labels)
  RDL = reformat_DATLAB(data.eeg,data.labels,aggregateperid=TRUE)
  DAT = RDL$DAT
  LAB = RDL$LAB
  P = split_data(LAB,DAT,logfile = logfile,proto_i=1,split=c(2,2),
                 uselog = FALSE,logdur=10)
  LABval = P$LABval;LABtest=P$LABtest;LABtrain=P$LABtrain
  DATval=P$DATval;DATtest=P$DATtest;DATtrain=P$DATtrain
  featuredict = create_featuredict(DAT)
  mymodel = train_model(DATtrain,LABtrain,DATval,LABval,featuredict,classifier="rf")

## End(Not run)

NLeSC/EEG-epilepsy-diagnosis documentation built on May 7, 2019, 6:02 p.m.