/**
* Sets up the panel with a new set of instances, attempting to guess the correct settings for
* various columns.
*
* @param newInstances the new set of results.
*/
public void setInstances(Instances newInstances) {
m_Instances = newInstances;
m_TTester.setInstances(m_Instances);
m_FromLab.setText("Got " + m_Instances.numInstances() + " results");
// Temporarily remove the configuration listener
m_RunCombo.removeActionListener(m_ConfigureListener);
// Do other stuff
m_DatasetKeyModel.removeAllElements();
m_RunModel.removeAllElements();
m_ResultKeyModel.removeAllElements();
m_CompareModel.removeAllElements();
int datasetCol = -1;
int runCol = -1;
String selectedList = "";
String selectedListDataset = "";
// =============== BEGIN EDIT melville ===============
boolean noise = false; // keep track of whether noise levels eval is required
boolean learning = false; // keep track of whether learning curve eval is required
boolean fraction =
false; // keep track of whether fractions of datasets are provided for learning
// the key on which to base the learning curves (either total instances or fraction)
int learning_key = -1;
boolean classificationTask =
false; // used to determine if regression measures should be selected
// =============== END EDIT melville ===============
for (int i = 0; i < m_Instances.numAttributes(); i++) {
String name = m_Instances.attribute(i).name();
m_DatasetKeyModel.addElement(name);
m_RunModel.addElement(name);
m_ResultKeyModel.addElement(name);
m_CompareModel.addElement(name);
// =============== BEGIN EDIT melville ===============
// If learning curves were generated then treat each
// dataset + pt combination as a different dataset
if (name.toLowerCase().equals("key_noise_levels")) {
// noise overrides learning curves - but treat noise levels
// like pts on learning curve
learning_key = i;
learning = true;
noise = true;
// fraction = true;
} else if (name.toLowerCase().equals("key_fraction_instances") && !noise) {
// fraction overrides total_instances
learning_key = i;
learning = true;
fraction = true;
} else if (name.toLowerCase().equals("key_total_instances") && !learning) {
learning_key = i;
learning = true;
} else
// =============== END EDIT melville ===============
if (name.toLowerCase().equals("key_dataset")) {
m_DatasetKeyList.addSelectionInterval(i, i);
selectedListDataset += "," + (i + 1);
} else if ((runCol == -1) && (name.toLowerCase().equals("key_run"))) {
m_RunCombo.setSelectedIndex(i);
runCol = i;
} else if (name.toLowerCase().equals("key_scheme")
|| name.toLowerCase().equals("key_scheme_options")
|| name.toLowerCase().equals("key_scheme_version_id")) {
m_ResultKeyList.addSelectionInterval(i, i);
selectedList += "," + (i + 1);
// =============== BEGIN EDIT mbilenko ===============
// automatic selection of the correct measure for clustering experiments
} else if (name.toLowerCase().indexOf("pairwise_f_measure") != -1) {
m_CompareCombo.setSelectedIndex(i);
m_ErrorCompareCol = i;
}
// automatic selection of the correct measure for deduping experiments
else if (name.toLowerCase().equals("precision")) {
m_CompareCombo.setSelectedIndex(i);
// =============== END EDIT mbilenko ===============
} else if (name.toLowerCase().indexOf("percent_correct") != -1) {
m_CompareCombo.setSelectedIndex(i);
classificationTask = true;
} else if (!classificationTask
&& (name.toLowerCase().indexOf("root_mean_squared_error") != -1)) {
// automatic selection of the correct measure for regression experiments
m_CompareCombo.setSelectedIndex(i);
} else if (name.toLowerCase().indexOf("percent_incorrect") != -1) {
m_ErrorCompareCol = i;
// remember index of error for computing error reductions
}
}
// =============== BEGIN EDIT melville ===============
if (learning) {
m_DatasetKeyList.addSelectionInterval(learning_key, learning_key);
selectedListDataset += "," + (learning_key + 1);
m_CompareModel.addElement("%Error_reduction");
m_CompareModel.addElement("Top_20%_%Error_reduction");
}
// =============== END EDIT melville ===============
if (runCol == -1) {
runCol = 0;
}
m_DatasetKeyBut.setEnabled(true);
m_RunCombo.setEnabled(true);
m_ResultKeyBut.setEnabled(true);
m_CompareCombo.setEnabled(true);
// Reconnect the configuration listener
m_RunCombo.addActionListener(m_ConfigureListener);
// Set up the TTester with the new data
m_TTester.setRunColumn(runCol);
Range generatorRange = new Range();
if (selectedList.length() != 0) {
try {
generatorRange.setRanges(selectedList);
} catch (Exception ex) {
ex.printStackTrace();
System.err.println(ex.getMessage());
}
}
m_TTester.setResultsetKeyColumns(generatorRange);
generatorRange = new Range();
if (selectedListDataset.length() != 0) {
try {
generatorRange.setRanges(selectedListDataset);
} catch (Exception ex) {
ex.printStackTrace();
System.err.println(ex.getMessage());
}
}
m_TTester.setDatasetKeyColumns(generatorRange);
// =============== BEGIN EDIT melville ===============
m_TTester.setLearningCurve(learning);
m_TTester.setFraction(fraction);
if (learning) { // get points on the learning curve
Attribute attr;
if (noise) { // override fraction
attr = m_Instances.attribute("Key_Noise_levels");
} else if (fraction) {
attr = m_Instances.attribute("Key_Fraction_instances");
} else {
attr = m_Instances.attribute("Key_Total_instances");
}
double[] pts = new double[attr.numValues()];
for (int k = 0; k < attr.numValues(); k++) {
pts[k] = Double.parseDouble(attr.value(k));
}
// sort points
Arrays.sort(pts);
m_TTester.setPoints(pts);
}
// =============== END EDIT melville ===============
m_SigTex.setEnabled(true);
m_PrecTex.setEnabled(true);
setTTester();
}
public void actionPerformed(ActionEvent e) {
m_TTester.setRunColumn(m_RunCombo.getSelectedIndex());
setTTester();
}