public double GetHeadC45Scale(TrainingSet set)
{
TrainingSet ordered = DoOrderC45(set);
double maxC45scale = 0.0d;
for (int i = 0; i < ordered.Samples.Count; i++)
{
int c45 = ordered.Samples[i].C45;
double scale = GetC45EarningScale(ordered, c45);
if (i == 0)
{
maxC45scale = scale;
}
if (maxC45scale < scale)
{
maxC45scale = scale;
}
}
return(maxC45scale);
}
static void Main(string[] args)
{
TrainingEntry entry1 = new TrainingEntry("AGE", null);
TrainingEntry entry2 = new TrainingEntry("MODEL", new string[] { "X3", "X5" });
TrainingEntry entry3 = new TrainingEntry("GENDER", new string[] { "MALE", "FEMALE" });
TrainingSet set = new TrainingSet("SATISFIED", entry1, entry2, entry3);
set.AddSample(new TrainingSample(false, 21, "X5", "MALE"));
set.AddSample(new TrainingSample(true, 19, "X3", "FEMALE"));
set.AddSample(new TrainingSample(false, 22, "X5", "MALE"));
set.AddSample(new TrainingSample(true, 21, "X3", "MALE"));
set.AddSample(new TrainingSample(true, 30, "X3", "MALE"));
set.AddSample(new TrainingSample(false, 60, "X3", "FEMALE"));
set.AddSample(new TrainingSample(false, 45, "X3", "FEMALE"));
set.AddSample(new TrainingSample(false, 55, "X3", "MALE"));
set.Lock();
Trainer trainer = new Trainer(set);
trainer.TrainID3(set);
Console.ReadKey();
}
public TrainingSet(TrainingSet other, List <TrainingSample> samples)
{
this.OutputName = other.OutputName;
this.Entries = other.Entries;
this.Samples = samples;
}
public TrainingSet(TrainingSet other)
{
this.OutputName = other.OutputName;
this.Entries = other.Entries;
this.Samples = other.Samples;
}
public bool CalculateResult(TrainingSet set, string attibrute)
{
int index = GetEntryIDFor(set, attibrute);
int totalEntries = set.Entries.Length;
int totalOutCount = set.Entries[index].Values.Length;
string[] entryValues = set.Entries[index].Values;
string name = set.Entries[index].Name;
double scale = this.GetBranchEarningScale(set, index);
Dictionary <string, int> outCountsTrue = new Dictionary <string, int>(totalOutCount);
Dictionary <string, int> outCountsFalse = new Dictionary <string, int>(totalOutCount);
for (int i = 0; i < set.Samples.Count; i++)
{
string currentSample = set.Samples[i].Samples[index];
bool including = entryValues.Any(s => s.Equals(currentSample, StringComparison.InvariantCultureIgnoreCase));
if (including)
{
if (set.Samples[i].Output)
{
if (!outCountsTrue.ContainsKey(currentSample))
{
outCountsTrue.Add(currentSample, 1);
}
else
{
outCountsTrue[currentSample]++;
}
}
else
{
if (!outCountsFalse.ContainsKey(currentSample))
{
outCountsFalse.Add(currentSample, 1);
}
else
{
outCountsFalse[currentSample]++;
}
}
}
}
//Console.WriteLine("================================");
//Console.WriteLine($"NAME = {name}, SCALE = {scale}");
//Console.WriteLine("TRUE : " + outCountsTrue.Count);
//Console.WriteLine("FALSE : " + outCountsFalse.Count);
//if (outCountsTrue.ContainsKey(attibrute)) {
// Console.WriteLine("TRUE COUNT : " + outCountsTrue[attibrute]);
//}
//if (outCountsFalse.ContainsKey(attibrute)) {
// Console.WriteLine("FALSE COUNT : " + outCountsFalse[attibrute]);
//}
//Console.WriteLine("================================");
if (!outCountsTrue.ContainsKey(attibrute) && outCountsFalse.ContainsKey(attibrute))
{
return(false);
}
else if (outCountsTrue.ContainsKey(attibrute) && !outCountsFalse.ContainsKey(attibrute))
{
return(true);
}
return(false);
}
public double CalculateH(TrainingSet set)
{
return(this.CalculateH(set.GetP1(), set.GetP2()));
}
public void TrainID3(TrainingSet set)
{
string master = this.CalculateHead(set);
Console.WriteLine("HEAD NODE = " + master);
Console.WriteLine();
StringBuilder sb = new StringBuilder(1000);
for (int i = 0; i < set.Entries.Length; i++)
{
TrainingEntry current = set.Entries[i];
if (i == 0)
{
sb.AppendLine(GetHeadC45Name(set) + " <= " + GetHeadC45(set));
}
if (current.Name.Equals(master, StringComparison.InvariantCultureIgnoreCase))
{
for (int j = 0; j < current.Values.Length; j++)
{
string currentName = current.Values[j];
TrainingSet currentExcept = this.GetExceptBranch(set, currentName);
TrainingSet currentExcept2 = GetExceptLowerBranch(currentExcept, GetHeadC45(set));
string subBranch = CalculateHead(currentExcept2);
string subName = currentName;
sb.AppendLine("--------- SUB NODE = " + subBranch);
sb.AppendLine("\t\t VALUE = " + subName);
if (!subBranch.Equals(master))
{
sb.AppendLine("|");
sb.AppendLine("|");
sb.AppendLine("--------- SUB NODE = " + currentName);
PrintSubBranch(currentExcept2, sb, subBranch);
sb.AppendLine("|");
sb.AppendLine("|");
}
else
{
sb.AppendLine("\t\t(" + CalculateResult(currentExcept2, subName) + ")");
}
//Console.WriteLine("name = " + currentName + "\t\t sub = " + subName + " -- scale = " + currentScale);
}
}
if (i == 0)
{
sb.AppendLine(GetHeadC45Name(set) + " > " + GetHeadC45(set));
}
if (i == 0)
{
sb.AppendLine("(False)");
}
}
Console.WriteLine(sb.ToString());
}
public double GetBranchEarningScale(TrainingSet set, int index)
{
int totalEntries = set.Entries.Length;
//int totalOutCount = set.Entries[index].Values.Length;
string[] entryValues = set.Entries[index].Values;
if (entryValues == null && index == 0)
{
entryValues = set.Entries[1].Values;
}
//ISI soğuk, ılık, sıcak --- 4, 6, 4
Dictionary <string, int> outCountsTrue = new Dictionary <string, int>();
Dictionary <string, int> outCountsFalse = new Dictionary <string, int>();
for (int i = 0; i < set.Samples.Count; i++)
{
string currentSample = set.Samples[i].Samples[index];
if (set.Samples[i].Output)
{
if (!outCountsTrue.ContainsKey(currentSample))
{
outCountsTrue.Add(currentSample, 1);
}
else
{
outCountsTrue[currentSample]++;
}
}
else
{
if (!outCountsFalse.ContainsKey(currentSample))
{
outCountsFalse.Add(currentSample, 1);
}
else
{
outCountsFalse[currentSample]++;
}
}
}
double totalH = 0.0d;
//Yüksek olanı al ?
if (outCountsTrue.Count > 0 && outCountsFalse.Count == 0)
{
return(1.0d);
}
else if (outCountsTrue.Count == 0 && outCountsFalse.Count > 0)
{
return(0.0d);
}
foreach (KeyValuePair <string, int> kvp in outCountsTrue)
{
int countTrue = 0;
int countFalse = 0;
if (outCountsTrue.ContainsKey(kvp.Key))
{
countTrue = outCountsTrue[kvp.Key];
}
if (outCountsFalse.ContainsKey(kvp.Key))
{
countFalse = outCountsFalse[kvp.Key];
}
int countTotal = countTrue + countFalse;
double currentH = 0.0d;
if (countTrue == 0 && countFalse > 0)
{
currentH = this.CalculateH(countFalse / (double)countTotal, countFalse / (double)countTotal);
}
else if (countTrue > 0 && countFalse > 0)
{
currentH = this.CalculateH(countTrue / (double)countTotal, countFalse / (double)countTotal);
}
else if (countTrue > 0 && countFalse == 0)
{
currentH = this.CalculateH(countTrue / (double)countTotal, countTrue / (double)countTotal);
}
totalH += (double)((double)((double)countTotal / (double)set.GetDataCount()) * currentH);
//Console.WriteLine("Key = {0}, Value = {1}, TRUE = {2}, FALSE = {3}, H = {4}", kvp.Key, kvp.Value, countTrue, countFalse, currentH);
}
return(this.CalculateH(set) - totalH);
}
public Trainer(TrainingSet set)
{
this.m_set = set;
this.m_results = new List <EntryResult>();
}
