use of org.knime.base.node.viz.enrichment2.EnrichmentPlotterSettings.Curve in project knime-core by knime.
the class EnrichmentPlotterModel method execute.
/**
* {@inheritDoc}
*/
@Override
protected BufferedDataTable[] execute(final BufferedDataTable[] inData, final ExecutionContext exec) throws Exception {
final double rowCount = inData[0].size();
final BufferedDataContainer areaOutCont = exec.createDataContainer(AREA_OUT_SPEC);
final BufferedDataContainer discrateOutCont = exec.createDataContainer(getDiscrateOutSpec());
final double[] fractionSizes = m_settings.getFractionSizes();
for (int i = 0; i < m_settings.getCurveCount(); i++) {
final ExecutionMonitor sexec = exec.createSubProgress(1.0 / m_settings.getCurveCount());
exec.setMessage("Generating curve " + (i + 1));
final Curve c = m_settings.getCurve(i);
final Helper[] curve = new Helper[KnowsRowCountTable.checkRowCount(inData[0].size())];
final int sortIndex = inData[0].getDataTableSpec().findColumnIndex(c.getSortColumn());
final int actIndex = inData[0].getDataTableSpec().findColumnIndex(c.getActivityColumn());
int k = 0, maxK = 0;
for (DataRow row : inData[0]) {
DataCell c1 = row.getCell(sortIndex);
DataCell c2 = row.getCell(actIndex);
if (k++ % 100 == 0) {
sexec.checkCanceled();
sexec.setProgress(k / rowCount);
}
if (c1.isMissing()) {
continue;
} else {
curve[maxK] = new Helper(((DoubleValue) c1).getDoubleValue(), c2);
}
maxK++;
}
Arrays.sort(curve, 0, maxK);
if (c.isSortDescending()) {
for (int j = 0; j < maxK / 2; j++) {
Helper h = curve[j];
curve[j] = curve[maxK - j - 1];
curve[maxK - j - 1] = h;
}
}
// this is for down-sampling so that the view is faster;
// plotting >100,000 points takes quite a long time
final int size = Math.min(MAX_RESOLUTION, maxK);
final double downSampleRate = maxK / (double) size;
final double[] xValues = new double[size + 1];
final double[] yValues = new double[size + 1];
xValues[0] = 0;
yValues[0] = 0;
int lastK = 0;
double y = 0, area = 0;
int nextHitRatePoint = 0;
final double[] hitRateValues = new double[fractionSizes.length];
final HashMap<DataCell, MutableInteger> clusters = new HashMap<DataCell, MutableInteger>();
// set hit rate values for fractions that are smaller than 1 row to 0
while ((maxK * fractionSizes[nextHitRatePoint] / 100) < 1) {
hitRateValues[nextHitRatePoint++] = 0;
}
for (k = 1; k <= maxK; k++) {
final Helper h = curve[k - 1];
if (m_settings.plotMode() == PlotMode.PlotSum) {
y += ((DoubleValue) h.b).getDoubleValue();
} else if (m_settings.plotMode() == PlotMode.PlotHits) {
if (!h.b.isMissing() && (((DoubleValue) h.b).getDoubleValue() >= m_settings.hitThreshold())) {
y++;
}
} else if (!h.b.isMissing()) {
MutableInteger count = clusters.get(h.b);
if (count == null) {
count = new MutableInteger(0);
clusters.put(h.b, count);
}
if (count.inc() == m_settings.minClusterMembers()) {
y++;
}
}
area += y / maxK;
if ((int) (k / downSampleRate) >= lastK + 1) {
lastK++;
xValues[lastK] = k;
yValues[lastK] = y;
}
// thats why this needs to be a while
while ((nextHitRatePoint < fractionSizes.length) && (k == (int) Math.floor(maxK * fractionSizes[nextHitRatePoint] / 100))) {
hitRateValues[nextHitRatePoint] = y;
nextHitRatePoint++;
}
}
xValues[xValues.length - 1] = maxK;
yValues[yValues.length - 1] = y;
area /= y;
m_curves.add(new EnrichmentPlot(c.getSortColumn() + " vs " + c.getActivityColumn(), xValues, yValues, area));
areaOutCont.addRowToTable(new DefaultRow(new RowKey(c.toString()), new DoubleCell(area)));
for (int j = 0; j < hitRateValues.length; j++) {
hitRateValues[j] /= y;
}
double[] enrichmentFactors = new double[hitRateValues.length];
for (int j = 0; j < enrichmentFactors.length; j++) {
enrichmentFactors[j] = calculateEnrichmentFactor(hitRateValues[j], fractionSizes[j]);
}
discrateOutCont.addRowToTable(new DefaultRow(new RowKey(c.toString()), ArrayUtils.addAll(hitRateValues, enrichmentFactors)));
}
areaOutCont.close();
discrateOutCont.close();
return new BufferedDataTable[] { areaOutCont.getTable(), discrateOutCont.getTable() };
}
Aggregations