private void ButtonClassify_Click(object sender, EventArgs e)
{
var start = DateTime.Now;
if (CBRandomOrder.Checked)
{
SetDataRandomOrder();
}
classifier.SetClass(SelectClass.SelectedItem.ToString());
classifier.LoadData(data);
classifier.TrainModel();
TextOutput.ResetText();
foreach (var row in TextTestData.Text.Split('\n'))
{
TextOutput.AppendText($"{row}\t=> {classifier.TrainedModel.GetEstimatedClass(row)}\n");
}
TextOutput.AppendText($"\nExecute time:\t{new TimeSpan(DateTime.Now.Ticks - start.Ticks).TotalSeconds.ToString()}s\n");
}
private void ButtonGenerateModel_Click(object sender, EventArgs e)
{
var start = DateTime.Now;
if (CBRandomOrder.Checked)
{
SetDataRandomOrder();
}
classifier.SetClass(SelectClass.SelectedItem.ToString());
int[][] confusionMatrix = null;
int dataCount = data.Count();
switch (SelectTestOption.SelectedIndex)
{
//Use Test Set
case 0:
classifier.LoadData(data);
classifier.TrainModel();
confusionMatrix = classifier.GetConfusionMatrix(TextTestData.Text);
break;
//Cross-Validation
case 1:
int fold = int.Parse(TextFold.Text);
int foldCount = dataCount / fold;
for (int j = 0; j < fold; j++)
{
int c = foldCount * j;
classifier.LoadData(data.Take(c).Concat(data.Skip(c + foldCount)));
classifier.TrainModel();
confusionMatrix = AddMatrix(confusionMatrix,
classifier.GetConfusionMatrix(data.Skip(c).Take(foldCount)));
}
break;
//Percentage Split
case 2:
dataCount = dataCount * int.Parse(TextPercentage.Text) / 100;
classifier.LoadData(data.Take(dataCount));
classifier.TrainModel();
confusionMatrix = classifier.GetConfusionMatrix(data.Skip(dataCount));
break;
default:
confusionMatrix = new int[0][];
break;
}
TextOutput.ResetText();
if (CBConfusionMatrix.Checked)
{
TextOutput.AppendText("Confusion Matrix:\n");
TextOutput.AppendText($"\t{string.Join("\t", classifier.ClassInfo.NominalOption)}\n");
var classOption = classifier.ClassInfo.NominalOption.GetEnumerator();
for (int j = 0; j < classifier.ClassInfo.NominalOptionCount; j++)
{
classOption.MoveNext();
TextOutput.AppendText($"{classOption.Current}\t{string.Join("\t", confusionMatrix[j])}\n");
}
TextOutput.AppendText("\n");
}
var itor = Enumerable.Range(0, confusionMatrix.Length);
decimal testCount = confusionMatrix.SelectMany(x => x).Sum();
TextOutput.AppendText($"Accuracy:\t{(itor.Select(x => confusionMatrix[x][x]).Sum() / testCount).ToString()}\n\n");
int i = 0;
decimal d;
foreach (var o in classifier.ClassInfo.NominalOption)
{
TextOutput.AppendText($"{o}：\n");
TextOutput.AppendText("Precision:\t");
d = itor.Select(x => confusionMatrix[x][i]).Sum();
TextOutput.AppendText(d == 0m ? "∞" : (confusionMatrix[i][i] / d).ToString());
TextOutput.AppendText("\n");
d = itor.Select(x => confusionMatrix[i][x]).Sum();
TextOutput.AppendText("Recall:\t");
TextOutput.AppendText(d == 0m ? "∞" : (confusionMatrix[i][i] / d).ToString());
TextOutput.AppendText("\n");
d = itor.Select(x => confusionMatrix[i][x]).Concat(
itor.Select(x => confusionMatrix[x][i])).Sum();
TextOutput.AppendText("F1-measure:\t");
TextOutput.AppendText(d == 0m ? "∞" : (confusionMatrix[i][i] * 2 / d).ToString());
TextOutput.AppendText("\n\n");
i++;
}
TextOutput.AppendText($"Execute time:\t{new TimeSpan(DateTime.Now.Ticks - start.Ticks).TotalSeconds.ToString()}s\n");
}
