public ID3Node GetID3Node(IEnumerable<Tuple<List<String>, String>> data, List<int> UsedFeatures)
{
ID3Node node = new ID3Node();
node.TargetClassProbabilities = TargetClassProbabilities(data);
node.Entropy = Entropy(node.TargetClassProbabilities);
if (node.Entropy > 0)
{
node.DecisionVariableIndex = GetDecisionVariable(node, data, UsedFeatures);
if (node.DecisionVariableIndex != -1)
{
List<String> DecisionVariableValues = ((from tuple in data
select tuple.Item1[node.DecisionVariableIndex]).Distinct()).ToList<String>();
if (DecisionVariableValues.Count > 1)
{
node.Children = new Dictionary<string, ID3Node>();
UsedFeatures.Add(node.DecisionVariableIndex);
foreach (String DecisionVariableValue in DecisionVariableValues)
{
var subset = from tuple in data
where tuple.Item1[node.DecisionVariableIndex] == DecisionVariableValue
select tuple;
node.Children[DecisionVariableValue] = GetID3Node(subset,UsedFeatures);
}
UsedFeatures.Remove(node.DecisionVariableIndex);
}
}
}
return node;
}
public int GetDecisionVariable(ID3Node node, IEnumerable<Tuple<List<String>, String>> data, List<int> UsedFeatures)
{
int NumFeatures = data.First().Item1.Count;
if (UsedFeatures.Count < NumFeatures)
{
Dictionary<int, double> Gain = new Dictionary<int, double>();
for (int i = 0; i < NumFeatures; i++)
{
if (UsedFeatures.Contains(i))
continue;
Gain[i] = InformationGain(node.Entropy, data, i);
}
double MaxGain = Gain.Values.Max();
foreach (int i in Gain.Keys)
{
if (Gain[i] == MaxGain)
return i;
}
}
return -1;
}
public int GetDecisionVariable(ID3Node node, IEnumerable <Tuple <List <String>, String> > data, List <int> UsedFeatures)
{
int NumFeatures = data.First().Item1.Count;
if (UsedFeatures.Count < NumFeatures)
{
Dictionary <int, double> Gain = new Dictionary <int, double>();
for (int i = 0; i < NumFeatures; i++)
{
if (UsedFeatures.Contains(i))
{
continue;
}
Gain[i] = InformationGain(node.Entropy, data, i);
}
double MaxGain = Gain.Values.Max();
foreach (int i in Gain.Keys)
{
if (Gain[i] == MaxGain)
{
return(i);
}
}
}
return(-1);
}
public ID3Node GetID3Node(IEnumerable <Tuple <List <String>, String> > data, List <int> UsedFeatures)
{
ID3Node node = new ID3Node();
node.TargetClassProbabilities = TargetClassProbabilities(data);
node.Entropy = Entropy(node.TargetClassProbabilities);
if (node.Entropy > 0)
{
node.DecisionVariableIndex = GetDecisionVariable(node, data, UsedFeatures);
if (node.DecisionVariableIndex != -1)
{
List <String> DecisionVariableValues = ((from tuple in data
select tuple.Item1[node.DecisionVariableIndex]).Distinct()).ToList <String>();
if (DecisionVariableValues.Count > 1)
{
node.Children = new Dictionary <string, ID3Node>();
UsedFeatures.Add(node.DecisionVariableIndex);
foreach (String DecisionVariableValue in DecisionVariableValues)
{
var subset = from tuple in data
where tuple.Item1[node.DecisionVariableIndex] == DecisionVariableValue
select tuple;
node.Children[DecisionVariableValue] = GetID3Node(subset, UsedFeatures);
}
UsedFeatures.Remove(node.DecisionVariableIndex);
}
}
}
return(node);
}
public void Train(String TrainingDataFile, char FieldSeparator, bool ContainsHeader, int TargetVariablePositionZeroBased)
{
IEnumerable <Tuple <List <String>, String> > data =
from lines in File.ReadAllLines(TrainingDataFile)
.Skip(ContainsHeader ? 1 : 0)
let columns = lines.Split(new[] { FieldSeparator }).ToList <String>()
let target = columns[TargetVariablePositionZeroBased]
let features = GetFeaturesAfterPruning(columns, TargetVariablePositionZeroBased)
select new Tuple <List <String>, String>(features, target);
List <int> UsedFeatures = new List <int>();
Root = GetID3Node(data, UsedFeatures);
}
public void Train(String TrainingDataFile, char FieldSeparator, bool ContainsHeader, int TargetVariablePositionZeroBased)
{
IEnumerable<Tuple<List<String>, String>> data =
from lines in File.ReadAllLines(TrainingDataFile)
.Skip(ContainsHeader ? 1 : 0)
let columns = lines.Split(new[] { FieldSeparator }).ToList<String>()
let target = columns[TargetVariablePositionZeroBased]
let features = GetFeaturesAfterPruning(columns, TargetVariablePositionZeroBased)
select new Tuple<List<String>, String>(features, target);
List<int> UsedFeatures = new List<int>();
Root = GetID3Node(data, UsedFeatures);
}