// Called when the user tries to train the classifier with the data previously loaded.
private void trainClassifier_button_Click(object sender, EventArgs e)
{
// Scroll to the end of the text box.
classificationLog_richTextBox.SelectionStart = classificationLog_richTextBox.Text.Length;
classificationLog_richTextBox.ScrollToCaret();
// Return if there was an error when processing the training dataset.
if (!trainingData.ProcessDataset(attributeToPredict_comboBox.SelectedItem.ToString()))
{
classificationLog_richTextBox.SelectionColor = Color.Red;
classificationLog_richTextBox.SelectedText =
"=====> There was a problem with processing the training data." +
Environment.NewLine;
return;
}
classificationLog_richTextBox.SelectionColor = Color.Green;
classificationLog_richTextBox.SelectedText =
"=====> Training data was processed correctly." +
Environment.NewLine;
// Choose classifier to use.
switch (classifierToUse_comboBox.SelectedItem.ToString())
{
case "Decision Tree":
classifier = new DecisionTreeClassifier();
// Give the inputs' names to the classifier,
// otherwise they will be generated automatically.
((DecisionTreeClassifier)classifier).DecisionVariableNames =
trainingData.InputColumnNames;
break;
case "Naive Bayesian":
classifier = new NaiveBayesianClassifier();
break;
case "SVM":
classifier = new SVMClassifier();
break;
default:
// Return if no valid classifier is selected.
return;
}
// Measure and display training time and error.
chronometer.Reset();
chronometer.Start();
try
{
if (classifier is DecisionTreeClassifier)
{
classifierError = ((DecisionTreeClassifier)classifier).TrainClassifierWithParameters(
trainingData,
(int)treeJoin_numericUpDown.Value,
(int)treeHeight_numericUpDown.Value);
}
else if (classifier is NaiveBayesianClassifier)
{
classifierError = ((NaiveBayesianClassifier)classifier).TrainClassifier(
trainingData);
}
else if (classifier is SVMClassifier)
{
classifierError = ((SVMClassifier)classifier).TrainClassifierWithParameters(
trainingData,
kernelList[svmKernel_comboBox.SelectedValue.ToString()],
/*svmAlgorithm_comboBox*/ null);
}
}
catch
{
classificationLog_richTextBox.SelectionColor = Color.Red;
classificationLog_richTextBox.SelectedText =
"=====> There was a problem with training the classifier." +
Environment.NewLine;
return;
}
chronometer.Stop();
classificationLog_richTextBox.SelectionColor = Color.Green;
classificationLog_richTextBox.SelectedText =
"=====> Classifier was trained successfully." +
Environment.NewLine;
trainingTimeValue_label.ForeColor = Color.Blue;
trainingTimeValue_label.Text = chronometer.ElapsedMilliseconds + " ms";
classifierErrorValue_label.ForeColor = (classifierError == 0) ? Color.Green : Color.Red;
classifierErrorValue_label.Text =
string.Format("{0:0.00}", classifierError * 100) + "%";
dataTrained = true;
// Deactivate some window's controls if
// classifier's training was successful.
attributeToPredict_comboBox.Enabled = false;
classifierToUse_comboBox.Enabled = false;
trainingFile_button.Enabled = false;
trainClassifier_button.Enabled = false;
treeJoin_numericUpDown.Enabled = false;
treeHeight_numericUpDown.Enabled = false;
svmKernel_comboBox.Enabled = false;
svmAlgorithm_comboBox.Enabled = false;
decisionTree_checkBox.Enabled = false;
testClassifier_button.Enabled = trainingFileLoaded && testingFileLoaded && dataTrained;
this.Refresh();
// Show a visual tree if decision tree classifier
// was chosen and the user checked the check box.
if ((classifier is DecisionTreeClassifier) && decisionTree_checkBox.Checked)
{
try
{
VisualDecisionTree visualTree = new VisualDecisionTree(
((DecisionTreeClassifier)classifier).ClassificationDecisionTree);
Bitmap image = visualTree.drawHorizontalTree(
trainingData.OutputColumnName,
trainingData.CodeBook);
treeVisualizer = new TreeView(image);
}
catch
{
classificationLog_richTextBox.SelectionColor = Color.Red;
classificationLog_richTextBox.SelectedText =
"=====> There was a problem with drawing the decision tree." +
Environment.NewLine;
return;
}
treeVisualizer.Show();
}
}
// Called when the user tries to test the classifier with some testing data.
private void testClassifier_button_Click(object sender, EventArgs e)
{
// Scroll to the end of the text box.
classificationLog_richTextBox.SelectionStart = classificationLog_richTextBox.Text.Length;
classificationLog_richTextBox.ScrollToCaret();
// Return if there was an error when processing the testing dataset.
if (!testingData.ProcessDataset(trainingData.OutputColumnName, trainingData.CodeBook))
{
classificationLog_richTextBox.SelectionColor = Color.Red;
classificationLog_richTextBox.SelectedText =
"=====> There was a problem with processing the testing data." +
Environment.NewLine;
return;
}
classificationLog_richTextBox.SelectionColor = Color.Green;
classificationLog_richTextBox.SelectedText =
"=====> Testing data was processed correctly. Testing results:" +
Environment.NewLine;
// Reset prediction error.
predictionError = 0;
// Measure testing time.
chronometer.Reset();
chronometer.Start();
int[] predictedValues = classifier.TestClassifier(testingData);
chronometer.Stop();
// Log results and calculate prediction error.
for (int n = 0; n < predictedValues.Length; ++n)
{
// Verify if predicted results are equal to the actual results.
if (predictedValues[n] == testingData.OutputData[n])
{
classificationLog_richTextBox.SelectionColor = Color.Green;
classificationLog_richTextBox.SelectedText =
"MATCHING: expected " +
testingData.CodeBook.Translate(
testingData.OutputColumnName,
testingData.OutputData[n]) +
" and predicted " +
testingData.CodeBook.Translate(
testingData.OutputColumnName,
predictedValues[n]) +
Environment.NewLine;
}
else
{
++predictionError;
classificationLog_richTextBox.SelectionColor = Color.Red;
classificationLog_richTextBox.SelectedText =
"NOT MATCHNG: expected " +
testingData.CodeBook.Translate(
testingData.OutputColumnName,
testingData.OutputData[n]) +
" but predicted " +
testingData.CodeBook.Translate(
testingData.OutputColumnName,
predictedValues[n]) +
Environment.NewLine;
}
}
// Scroll to the end of the text box.
classificationLog_richTextBox.SelectionStart = classificationLog_richTextBox.Text.Length;
classificationLog_richTextBox.ScrollToCaret();
// Close the old window showing the confusion matrix
// (if present) and show a new window if the user
// checked the check box.
if (matrixVisualizer != null)
{
matrixVisualizer.Close();
}
if (confusionMatrix_checkBox.Checked)
{
matrixVisualizer = new ConfusionMatrixView(testingData, predictedValues);
matrixVisualizer.Show();
}
// Display testing time and prediction error.
testingTimeValue_label.ForeColor = Color.Blue;
testingTimeValue_label.Text = chronometer.ElapsedMilliseconds + " ms";
predictionErrorValue_label.ForeColor = (predictionError == 0) ? Color.Green : Color.Red;
predictionErrorValue_label.Text =
string.Format("{0:0.00}", predictionError * 100 / predictedValues.Length) + "%";
}