public List <DecisionTree> KFoldTest(int[][] X, int[] y, int K = 10, bool alert = false)
{
int KSampleSize = (int)Math.Ceiling(y.Length / (float)K);
List <DecisionTree> testTrees = new List <DecisionTree>();
int correctPredicts = 0;
int incorrectPredicts = 0;
for (int i = 0; i < K; i++)
{
var X_validation = X.Skip(i * KSampleSize).Take(KSampleSize).ToArray();
var y_validation = y.Skip(i * KSampleSize).Take(KSampleSize).ToArray();
var x_remain = X.Skip((i + 1) * KSampleSize);
var X_train = X.Take(i * KSampleSize).Concat(x_remain).ToArray();
var y_remain = y.Skip((i + 1) * KSampleSize);
var y_train = y.Take(i * KSampleSize).Concat(y_remain).ToArray();
var testTree = new DecisionTree();
testTree.Train(X_train, y_train);
testTrees.Add(testTree);
for (int j = 0; j < y_validation.Length; j++)
{
if (testTree.Predict(X_validation[j]) == y_validation[j])
{
correctPredicts++;
}
else
{
incorrectPredicts++;
}
}
}
var str = string.Format("Correct Predicts:{0} , IncorrectPredicts:{1} , avg acc:{2}, K-Num:{3}", correctPredicts, incorrectPredicts, correctPredicts / (float)(correctPredicts + incorrectPredicts), K);
Console.WriteLine(str);
if (alert)
{
MessageBox.Show(str, "Result");
}
return(testTrees);
}
private void Button_Click(object sender, RoutedEventArgs e)
{
var tag = int.Parse((KFolds.SelectedItem as ComboBoxItem).Tag.ToString());
if (tag == 0)
{
treeComboBox.Items.Clear();
var tree = new DecisionTree();
tree.Train(X, y);
trees = new List <DecisionTree>();
trees.Add(tree);
var item = new ComboBoxItem();
item.Content = "0";
item.Tag = "0";
treeComboBox.Items.Clear();
treeComboBox.Items.Add(item);
treeComboBox.SelectedIndex = 0;
}
else
{
int K = tag;
trees = KFoldTest(X, y, K, true);
treeComboBox.Items.Clear();
for (int i = 0; i < K; i++)
{
var item = new ComboBoxItem();
item.Content = i.ToString();
item.Tag = i.ToString();
treeComboBox.Items.Add(item);
}
treeComboBox.SelectedIndex = 0;
}
}
public MainWindow()
{
InitializeComponent();
GetData("divorce.csv", out X, out y, out string[] Xlabel, out string[] ylabel);
// veriseti randomizasyonu * kfold dan once
var bytes = new byte[y.Length];
new Random().NextBytes(bytes);
var bytesC = (byte[])bytes.Clone();
Array.Sort(bytes, X);
Array.Sort(bytesC, y);
KFoldTest(X, y, 20);
KFoldTest(X, y, 10);
KFoldTest(X, y, 5);
KFoldTest(X, y, 3);
KFoldTest(X, y, 2);
tree = new DecisionTree();
tree.Train(X, y);
questionList = new List <Question>();
for (int i = 0; i < Xlabel.Length; i++)
{
var q = new Question(i, Xlabel[i]);
questionList.Add(q);
xlist.Items.Add(q);
}
xtree.Items.Add(tree.rootNode);
}
