private void bFlatBtnExecute_Click_1(object sender, EventArgs e)
{
List <Attribute> attributes = createAttribute();
List <DecisionQuery> decision = createDecisionQuery();
if (decision.Count == 0 || attributes.Count == 0)
{
MessageBox.Show("File imported is empty!");
return;
}
DecisionQuery result = decision[decision.Count - 1];
decision.Remove(result);
int total = attributes.Count;
List <float> calcultaHS = new List <float>();
foreach (Answer question in result.possibleAnswers)
{
int count = 0;
foreach (Attribute attribute in attributes)
{
if (question.answer == attribute.answers[attribute.answers.Count - 1])
{
count++;
}
}
calcultaHS.Add(count);
}
hs = -calcultaHS[0] / total * (float)Math.Log(calcultaHS[0] / total) - calcultaHS[1] / total * (float)Math.Log(calcultaHS[1] / total);
richTextBox1.Text += "HS: " + -calcultaHS[0] + "/" + total + "*log(" + calcultaHS[0] + "/" + total + ")" + -calcultaHS[1] + "/" + total + "*log(" + calcultaHS[1] + "/" + total + ")=";
richTextBox1.Text += hs + "\n\n";
decisionTrees = new Node(null, decision, attributes);
levels = partition(decisionTrees, result, 1, String.Empty) + 1;
graph_Panel.Refresh();
return;
}
public Node(DecisionQuery partition_question, List <DecisionQuery> questions, List <Attribute> attributes)
{
this.partition_question = partition_question;
this.questions = questions;
this.attributes = attributes;
}
private int partition(Node parent, DecisionQuery result, int level, string possibleAnswerBranch)
{
if (parent.attributes.Count == 0 || parent.questions.Count == 0 || checkPartition(parent.attributes))
{
return(level);
}
List <List <List <KeyValuePair <float, float> > > > featurePairsVectors = new List <List <List <KeyValuePair <float, float> > > >();
List <List <List <float> > > featureVectors = computeFeatureVectors(parent.questions, parent.attributes, result, ref featurePairsVectors);
List <float> hsList = new List <float>();
for (int i = 0; i < featureVectors.Count; i++)
{
List <List <KeyValuePair <float, float> > > hs_question = new List <List <KeyValuePair <float, float> > >();
List <KeyValuePair <float, float> > hs_question1 = new List <KeyValuePair <float, float> >();
for (int j = 0; j < featureVectors[i].Count; j++)
{
richTextBox1.Text += "H(S" + possibleAnswerBranch + "_" + parent.questions[i].possibleAnswers[j].answer + "):";
float hs = 0;
for (int k = 0; k < featureVectors[i][j].Count; k++)
{
float hs1 = -featurePairsVectors[i][j][k].Key / featurePairsVectors[i][j][k].Value * (float)Math.Log(featurePairsVectors[i][j][k].Key / featurePairsVectors[i][j][k].Value);
richTextBox1.Text += -featurePairsVectors[i][j][k].Key + "/" + featurePairsVectors[i][j][k].Value;
richTextBox1.Text += "*log(" + featurePairsVectors[i][j][k].Key + "/" + featurePairsVectors[i][j][k].Value + ")";
if (featurePairsVectors[i][j][k].Key == 0)
{
hs1 = 0;
}
hs += hs1;
}
hs_question1.Add(new KeyValuePair <float, float>(hs, featurePairsVectors[i][j][0].Value));
richTextBox1.Text += "=" + hs + "\n";
}
float hs2 = 0;
float total = 0;
richTextBox1.Text += "H(S" + parent.questions[i].question + "):";
for (int t = 0; t < hs_question1.Count; t++)
{
total += hs_question1[t].Value;
}
for (int t = 0; t < hs_question1.Count; t++)
{
hs2 += hs_question1[t].Key * hs_question1[t].Value / total;
richTextBox1.Text += hs_question1[t].Value + "/" + total + "*" + hs_question1[t].Key;
if (t != hs_question1.Count - 1)
{
richTextBox1.Text += "+";
}
}
richTextBox1.Text += "=" + hs2 + "\n\n";
hsList.Add(hs2);
}
//int bestFeature = getBestFeature(featureVectors);
int bestFeature = getBestFeature(hsList);
richTextBox1.Text += " Ta chọn đặc tính tốt nhất: " + parent.questions[bestFeature].question.ToString() + "\n";
parent.partition_question = parent.questions[bestFeature];
List <DecisionQuery> brandClauses = new List <DecisionQuery>();
for (int i = 0; i < parent.questions.Count; i++)
{
if (i != bestFeature)
{
brandClauses.Add(parent.questions[i]);
}
}
List <List <Attribute> > brandsAttributes = new List <List <Attribute> >();
for (int i = 0; i < parent.questions[bestFeature].possibleAnswers.Count; i++)
{
List <Attribute> brandAttribute = new List <Attribute>();
foreach (Attribute attribute in parent.attributes)
{
if (attribute.answers[bestFeature] == parent.questions[bestFeature].possibleAnswers[i].answer)
{
Attribute temp = attribute.copy();
temp.answers.RemoveAt(bestFeature);
brandAttribute.Add(temp);
}
}
brandsAttributes.Add(brandAttribute);
}
parent.branch = new List <Node>();
int maxlv = level;
for (int i = 0; i < parent.questions[bestFeature].possibleAnswers.Count; i++)
{
// richTextBox1.Text += "-----Xét " + parent.questions[bestFeature].possibleAnswers.ToString() + "------";
Node brand = new Node(null, brandClauses, brandsAttributes[i]);
if (featureVectors[bestFeature][i].Max() != 1)
{
int lv = partition(brand, result, level + 1, parent.questions[bestFeature].possibleAnswers[i].answer);
if (lv > maxlv)
{
maxlv = lv;
}
}
parent.branch.Add(brand);
}
return(maxlv);
}
private void result_Algrithm(List <Attribute> attributes, DecisionQuery result, List <DecisionQuery> decisions)
{
float total = attributes.Count;
List <float> ratioHS = new List <float>();
for (int j = 0; j < result.possibleAnswers.Count; j++)
{
float count = 0;
for (int i = 0; i < attributes.Count; i++)
{
{
if (result.possibleAnswers[j].answer == attributes[i].answers[attributes[i].answers.Count - 1])
{
count++;
}
}
}
ratioHS.Add(count);
}
float HS = -ratioHS[0] / total * (float)Math.Log(ratioHS[0] / total) - ratioHS[1] / total * (float)Math.Log(ratioHS[1] / total);
richTextBox1.Text += "Entropy tại root node của bài toán là:\n";
richTextBox1.Text += "=> H(S)=" + -ratioHS[0] + "/" + total + ".log(" + ratioHS[0] + "/" + total + ")"
+ -ratioHS[1] + "/" + total + ".log(" + ratioHS[1] + "/" + total + ")=";
richTextBox1.Text += HS + "\n";
float Hmin = 0;
List <float> H = new List <float>();
int temp = 0;
for (int i = 0; i < decisions.Count; i++)
{
richTextBox1.Text += "Xét " + decisions[i].question.ToString() + "\n";
float Hsum = 0;
for (int j = 0; j < decisions[i].possibleAnswers.Count; j++)
{
float yes = decisions[i].possibleAnswers[j].yes;
float no = decisions[i].possibleAnswers[j].no;
float total1 = yes + no;
float HS1 = -yes / total1 * (float)Math.Log(yes / total1) - no / total1 * (float)Math.Log(no / total1);
if (yes == 0 || no == 0)
{
HS1 = 0;
}
richTextBox1.Text += "=> H(S" + decisions[i].possibleAnswers[j].answer + ")=" + -yes + "/" + total1 + ".log(" + yes + "/" + total1 + ")"
+ -no + "/" + total1 + ".log(" + no + "/" + total1 + ")=";
richTextBox1.Text += HS1 + "\n";
Hsum += HS1 * total1 / total;
}
H.Add(Hsum);
richTextBox1.Text += "==>H(" + decisions[i].question + ",S)=" + Hsum + "\n";
}
Hmin = H[0];
for (int k = 1; k < H.Count; k++)
{
if (H[k] < Hmin)
{
Hmin = H[k];
temp = k;
}
}
for (int i = 0; i < decisions.Count; i++)
{
if (i == temp)
{
richTextBox1.Text += "H(" + decisions[i].question + ",S) nhỏ nhất nên Gain lớn nhất." +
"==> Chọn " + decisions[i].question + " làm node gốc \n";
}
}
for (int i = 0; i < decisions[temp].possibleAnswers.Count; i++)
{
}
}
private List <List <List <float> > > computeFeatureVectors(List <DecisionQuery> decisionQueries, List <Attribute> attributes, DecisionQuery result, ref List <List <List <KeyValuePair <float, float> > > > featurePairsVectors)
{
List <List <List <float> > > featureVectors = new List <List <List <float> > >();
for (int c = 0; c < decisionQueries.Count; c++)
{
List <List <float> > featureVector = new List <List <float> >();
List <List <KeyValuePair <float, float> > > lKeyValuePairs = new List <List <KeyValuePair <float, float> > >();
for (int i = 0; i < decisionQueries[c].possibleAnswers.Count; i++)
{
List <float> vector = new List <float>();
List <KeyValuePair <float, float> > keyValue = new List <KeyValuePair <float, float> >();
for (int j = 0; j < result.possibleAnswers.Count; j++)
{
int count = 0;
int total = 0;
foreach (Attribute attribute in attributes)
{
if (attribute.answers[c] == decisionQueries[c].possibleAnswers[i].ANSWER)
{
total++;
if (attribute.answers[attribute.answers.Count - 1] == result.possibleAnswers[j].answer)
{
count++;
}
}
}
vector.Add((float)count / total);
keyValue.Add(new KeyValuePair <float, float>(count, total));
}
lKeyValuePairs.Add(keyValue);
featureVector.Add(vector);
//richTextBox1.Text += "H(S" + decisionQueries[c].possibleAnswers[i].answer +featureVector[i]..ToString()+ ")="+"\n";
}
featureVectors.Add(featureVector);
featurePairsVectors.Add(lKeyValuePairs);
}
return(featureVectors);
}