public IDataTable ConvertToNumeric(IDataTableVectoriser vectoriser = null, bool useTargetColumnIndex = true, Stream output = null)
{
var writer = new DataTableWriter(output);
vectoriser = vectoriser ?? GetVectoriser(useTargetColumnIndex);
// add the numeric columns
foreach (var name in vectoriser.ColumnNames)
{
writer.AddColumn(name, ColumnType.Float);
}
// add the classification label column
var classColumnIndex = TargetColumnIndex;
if (useTargetColumnIndex)
{
var classColumn = _column[classColumnIndex];
writer.AddColumn(classColumn.Name, ColumnType.String, true);
}
// vectorise each row
_Iterate((row, i) => {
var rowData = vectoriser.GetInput(row).Data.AsEnumerable().Cast <object>();
if (useTargetColumnIndex)
{
rowData = rowData.Concat(new object[] { row.GetField <string>(classColumnIndex) });
}
writer.AddRow(new DataTableRow(this, rowData.ToArray(), _rowConverter));
return(true);
});
return(writer.GetDataTable());
}
/// <summary>
/// Creates a data source from a data table
/// </summary>
/// <param name="dataTable">The data table to convert</param>
/// <param name="vectoriser">Optional data table vectoriser (if the data table contains categorical or index based data)</param>
/// <returns></returns>
public IDataSource CreateDataSource(IDataTable dataTable, IDataTableVectoriser vectoriser = null)
{
var columns = dataTable.Columns;
var dataColumnTypes = columns
.Where((c, i) => i != dataTable.TargetColumnIndex)
.Select(c => c.Type)
.ToList()
;
var targetColumnType = columns[dataTable.TargetColumnIndex].Type;
var firstDataColumnType = dataColumnTypes.FirstOrDefault();
if (firstDataColumnType != ColumnType.Null && dataColumnTypes.All(ct => ct == firstDataColumnType))
{
// many to many
if (firstDataColumnType == ColumnType.Matrix && targetColumnType == ColumnType.Matrix)
{
return(new SequentialDataTableAdaptor(_lap, dataTable));
}
// one to one
else if (firstDataColumnType == ColumnType.Vector && targetColumnType == ColumnType.Vector)
{
return(new VectorBasedDataTableAdaptor(_lap, dataTable));
}
// one to many
else if (firstDataColumnType == ColumnType.Vector && targetColumnType == ColumnType.Matrix)
{
return(new OneToManyDataTableAdaptor(_lap, dataTable));
}
// many to one
else if (firstDataColumnType == ColumnType.Matrix && targetColumnType == ColumnType.Vector)
{
return(new ManyToOneDataTableAdaptor(_lap, dataTable));
}
// volume classification
else if (firstDataColumnType == ColumnType.Tensor && targetColumnType == ColumnType.Vector)
{
return(new TensorBasedDataTableAdaptor(_lap, dataTable));
}
// index list
else if (firstDataColumnType == ColumnType.IndexList)
{
return(new IndexListDataTableAdaptor(_lap, dataTable, vectoriser));
}
// weighted index list
else if (firstDataColumnType == ColumnType.WeightedIndexList)
{
return(new WeightedIndexListDataTableAdaptor(_lap, dataTable, vectoriser));
}
}
// default adapator
return(new DefaultDataTableAdaptor(_lap, dataTable, vectoriser));
}
public DataTableTrainingDataProvider(ILinearAlgebraProvider lap, IDataTable table, IDataTableVectoriser vectoriser)
{
_lap = lap;
_table = table;
_vectoriser = vectoriser;
_inputSize = _vectoriser.InputSize;
_outputSize = _vectoriser.OutputSize;
}
public IDataTable ConvertToNumeric(IDataTableVectoriser vectoriser = null, Stream output = null)
{
var writer = new DataTableWriter(output);
vectoriser = vectoriser ?? GetVectoriser();
foreach (var name in vectoriser.ColumnNames)
{
writer.AddColumn(name, ColumnType.Float);
}
var classColumnIndex = TargetColumnIndex;
var classColumn = _column[classColumnIndex];
writer.AddColumn(classColumn.Name, ColumnType.String, true);
_Iterate(row => {
var data = vectoriser.GetInput(row).AsEnumerable().Cast <object>().Concat(new object[] { row.GetField <string>(classColumnIndex) });
writer.AddRow(data);
return(true);
});
return(writer.GetDataTable());
}
public DefaultDataTableAdaptor(ILinearAlgebraProvider lap, IDataTable dataTable, IDataTableVectoriser vectoriser = null)
: base(lap, dataTable)
{
_vectoriser = vectoriser ?? dataTable.GetVectoriser();
}
public ITrainingDataProvider CreateTrainingDataProvider(IDataTable table, IDataTableVectoriser vectoriser = null)
{
return(new DataTableTrainingDataProvider(_lap, table, vectoriser ?? table.GetVectoriser()));
}
