private double trainingDecision(DecisionTree decision)
{
double best = 0;
int index = 8;
for (int i = 2; i <= 9; i++)
{
decision.depthLimit = i;
decision.training(trainingSet);
double aux = Math.Round(decision.testing(testingSet), 2);
if (aux > best)
{
best = aux;
index = i;
}
advance++;
}
decision.depthLimit = index;
decision.training(trainingSet);
double result = Math.Round(decision.testing(testingSet), 2);
advance++;
return(result);
}
private void defineMessage(Node node, int number, DecisionTree decisionTreeParent)
{
string value = "";
switch (node.parent.value)
{
case 1:
if (number == 0)
{
value = "Age < 30";
}
else if (number == 1)
{
value = "Age < 60";
}
else
{
value = "Age >= 60";
}
break;
case 2:
if (number == 0)
{
value = "F";
}
else
{
value = "M";
}
break;
case 3:
value = "Type " + number;
break;
case 4:
if (number == 0)
{
value = "Blood pressure < " + decisionTreeParent.splitBloodPressure;
}
else
{
value = "Blood Pressure >= " + decisionTreeParent.splitBloodPressure;
}
break;
case 5:
if (number == 0)
{
value = "Cholesterol < " + decisionTreeParent.splitCholesterol;
}
else
{
value = "Cholesterol >= " + decisionTreeParent.splitCholesterol;
}
break;
case 6:
if (number == 0)
{
value = "Level sugar < 120 mg/dl";
}
else
{
value = "Level sugar >= 120 mg/dl";
}
break;
case 7:
if (number == 0)
{
value = "NO";
}
else
{
value = "YES";
}
break;
case 8:
value = "Value " + number;
break;
case 9:
if (number == 0)
{
value = "Heart rate max < " + decisionTreeParent.splitHeartRate;
}
else
{
value = "Heart rate max >= " + decisionTreeParent.splitHeartRate;
}
break;
}
node.message = value;
}
private void training()
{
bool haveAtributes = false;
for (int i = 1; i <= 9 && haveAtributes == false && trainingSet.Count > 0 && depth <= depthLimit; i++)
{
if (trainingSet.ElementAt(0).get(i) != -1)
{
haveAtributes = true;
}
}
if (haveAtributes)
{
mainEntropy = entropy();
if (mainEntropy == 0)
{
asnswer = trainingSet.ElementAt(0).result;
}
else
{
double[] gain = new double[9];
double max = -1;
int index = 0;
for (int i = 1; i <= 9; i++)
{
if (trainingSet.ElementAt(0).get(i) != -1)
{
if (!isCategoric(i))
{
splitContinuValues(i);
}
gain[i - 1] = mainEntropy - entropy(i);
if (max < gain[i - 1])
{
index = i;
max = gain[i - 1];
}
}
}
column = index;
int aux = 2;
if (column == 1 || column == 8)//variables de 3 valores
{
aux = 3;
}
else if (column == 3)
{
aux = 4;
}
childrenNode = new DecisionTree[aux];
List <Patient>[] all = new List <Patient> [aux];
for (int i = 0; i < all.Length; i++)
{
all[i] = new List <Patient>();
}
List <Patient> newPatients = new List <Patient>();
foreach (Patient patient in trainingSet)
{
int value = patient.get(column);
double safe = 0;
if (!isCategoric(column))
{
if (column == 4)
{
safe = splitBloodPressure;
}
else if (column == 5)
{
safe = splitCholesterol;
}
else if (column == 9)
{
safe = splitHeartRate;
}
if (patient.get(column) < safe)
{
value = 0;
}
else
{
value = 1;
}
}
if (column == 1)
{
if (value < 30)//Young
{
value = 0;
}
else if (value < 60)//Adult
{
value = 1;
}
else//Elder
{
value = 2;
}
}
int[] values = new int[11];
for (int j = 0; j < 11; j++)
{
values[j] = patient.get(j);
}
values[column] = -1;
Patient newPatient = new Patient(values[0], values[1], values[2], values[3], values[4], values[5], values[6], values[7], values[8], values[9], values[10]);
all[value].Add(newPatient);
}
if (column == 2)
{
}
for (int i = 0; i < childrenNode.Length; i++)
{
DecisionTree child = new DecisionTree();
child.depth = depth + 1;
child.depthLimit = depthLimit;
child.training(all[i]);
childrenNode[i] = child;
}
}
}
else
{
double ammountYes = 0;
double ammountNo = 0;
foreach (Patient patient in trainingSet)
{
if (patient.result == 0)
{
ammountNo++;
}
else
{
ammountYes++;
}
}
if (ammountYes / (ammountYes + ammountNo) > 0.54)
{
asnswer = 1;
}
else
{
asnswer = 0;
}
}
}
