public LinkedList <ResultRow> validate(ModelCreator modelcreator, Dictionary <string, ICollection <ICollection <Observation> > > dataSets, string gesture, int nSubsets)
{
var results = new LinkedList <ResultRow>();
foreach (var entry in dataSets)
{
var trueUser = entry.Key;
var trueUserData = entry.Value;
var subsetSize = trueUserData.Count / nSubsets;
var subsets = new LinkedList <ICollection <Observation> > [nSubsets];
var enu = trueUserData.GetEnumerator();
//fill subsets
for (int i = 0; i < nSubsets; i++)
{
subsets[i] = new LinkedList <ICollection <Observation> >();
for (int j = 0; j < subsetSize; j++)
{
enu.MoveNext();
subsets[i].AddLast(enu.Current);
}
}
int curSubset = 1;
foreach (var testSet in subsets)
{
//build trainSet
var trainSet = new LinkedList <ICollection <Observation> >();
foreach (var subSet in subsets)
{
if (subSet != testSet)
{
foreach (var trace in subSet)
{
trainSet.AddLast(trace);
}
}
}
//use trainSet to train the model
var model = modelcreator.createModel(trainSet);
//true positives, false negatives, false positives and true negatives
int TP = 0;
int FN = 0;
int FP = 0;
int TN = 0;
//calc estimated FRR
foreach (var testTrace in testSet)
{
if (model.evaluate(testTrace) == 0)
{
FN++;
}
else
{
TP++;
}
}
//double FRR = (double)nRejections / n;
//calc estimated FAR
foreach (var entry2 in dataSets)
{
var falseUser = entry2.Key;
var falseUserData = entry2.Value;
if (trueUser == falseUser)
{
continue;
}
foreach (var falseUserTrace in falseUserData)
{
if (model.evaluate(falseUserTrace) == 0)
{
TN++;
}
else
{
FP++;
}
}
}
//save result
var result = new ResultRow();
result.gesture = gesture;
result.user = trueUser;
result.subset = curSubset++;
result.FN = FN;
result.FP = FP;
result.TP = TP;
result.TN = TN;
results.AddLast(result);
}
}
return(results);
}
