/// <summary>
/// Returns a filled DataSet containing all valid information read in
/// from the *.names and *.data files.
/// </summary>
public DataSet BuildDataSet()
{
var dataSet = new DataSet();
BuildNamesEntries(ref dataSet);
BuildDataEntries(ref dataSet);
return dataSet;
}
/// <summary>
/// Returns a training/testing set split based on the given trainingSize.
/// </summary>
/// <param name="trainingSize">Number of data instances we want in our training set.</param>
/// <returns>List[DataSet] containing training/test sets</returns>
public List<DataSet> RandomInstance(double trainingSize)
{
DataSet instance = Shuffle();
var training = new DataSet { Features = new List<Feature>(this.Features) };
training.DataEntries.AddRange(instance.DataEntries.Take((int)trainingSize));
var test = new DataSet(instance.DataEntries.Except(training.DataEntries).ToList());
test.Features = new List<Feature>(this.Features);
training.OutputIndex = test.OutputIndex = this.OutputIndex;
return new List<DataSet>(){training, test};
}
/// <summary>
/// Shuffles this dataset and returns a new dataset containing the random instances.
/// </summary>
/// <returns>DataSet</returns>
private DataSet Shuffle()
{
var instance = new DataSet { Features = new List<Feature>(this.Features) };
for(int i=DataEntries.Count-1; i>=0; i--) {
int index = Rng.Next(DataEntries.Count);
while(instance.DataEntries.Contains(DataEntries[index]))
index = Rng.Next(DataEntries.Count);
instance.DataEntries.Add(DataEntries[index]);
}
return instance;
}
/// <summary>
/// Reads all valid entries(determined by DataReader) and stores values into a List[object]
/// in our DataSet; paired with the attributes read in from the names file.
/// </summary>
private void BuildDataEntries(ref DataSet dataSet)
{
foreach(string entry in m_DataReader.ValidEntries()) {
string[] data = entry.Split(',');
if(data.Length != dataSet.Features.Count){
Console.WriteLine("[Error]: Invalid # of data elements in {0}.", data.Select(s=>s.ToString(CultureInfo.InvariantCulture)));
continue;
}
var instance = new DataInstance();
for(int i=0; i<data.Length; i++) {
if(IsValidValue(dataSet.Features[i].Type, data[i], dataSet.Features[i].PossibleValues.ToArray()))
instance.Add(data[i]);
}
if(instance.Count == dataSet.Features.Count)
dataSet.DataEntries.Add(instance);
}
}
/// <summary>
/// Reads all valid entries (determined by NamesReader) and stores values into
/// and attribute object that is stored in our DataSet.
/// </summary>
private void BuildNamesEntries(ref DataSet dataSet)
{
foreach(string line in m_NamesReader.ValidEntries()) {
string[] features = line.Split(':');
var feature = new Feature(features[0], (Types)Enum.Parse(typeof(Types), features[1]));
foreach(string s in features[2].Split(',')) {
feature.PossibleValues.Add(s.Trim());
}
dataSet.Features.Add(feature);
if(feature.Type == Types.Output){
dataSet.OutputIndex = dataSet.Features.Count-1;
}
}
}
public KNearest(DataSet data)
{
m_DataSet = data;
}
