public Dataset(IDatasetReader <T> trainReader, IDatasetReader <T> testReader)
{
_trainSamples = trainReader.ReadAll().AsQueryable();
_testSamples = testReader.ReadAll().AsQueryable();
_evalSamples = Enumerable.Empty <T>().AsQueryable();
_allSamples = _trainSamples.Concat(_testSamples);
}
public Dataset(IDatasetReader <T> reader, Func <IEnumerable <T>, Tuple <IEnumerable <T>, IEnumerable <T>, IEnumerable <T> > > trainTestEvalSplitter)
{
_allSamples = reader.ReadAll().AsQueryable();
var splits = trainTestEvalSplitter(_allSamples);
_trainSamples = splits.Item1.AsQueryable();
_testSamples = splits.Item2.AsQueryable();
_evalSamples = splits.Item3.AsQueryable();
}
public Dataset(IDatasetReader <T> reader, double testPortion, double evalPortion)
{
if (testPortion + evalPortion > 1)
{
throw new Exception("Sum of test and evaluation portions should be less than 1.");
}
_allSamples = reader.ReadAll().AsQueryable();
if (testPortion == 0 && evalPortion == 0)
{
_trainSamples = _allSamples;
_testSamples = _evalSamples = Enumerable.Empty <T>().AsQueryable();
return;
}
int trainCount = Convert.ToInt32(_allSamples.Count() * (1 - testPortion - evalPortion));
int testCount = Convert.ToInt32(_allSamples.Count() * testPortion);
_trainSamples = _allSamples.Take(trainCount);
_testSamples = _allSamples.Skip(trainCount).Take(testCount);
_evalSamples = _allSamples.Skip(trainCount + testCount);
}
public override IEnumerable <ItemRatingWithRelations> ReadWithoutFiltering()
{
return(_itemRatingsReader.ReadAll().Select(ir => new ItemRatingWithRelations(ir, _relations)).ToList());
}