public void test_Interface(
String[] my_options, String[] my_data, int invalidLineCnt, int validLineCnt)
throws Exception {
try {
BufferedWriter out_csv = new BufferedWriter(new FileWriter(path_csv));
for (String aMy_data : my_data) {
out_csv.write(aMy_data + "\n");
}
out_csv.flush();
out_csv.close();
} catch (Exception e) {
System.err.print(e.getMessage());
}
CSVLoader.testMode = true;
CSVLoader.main(my_options);
// do the test
VoltTable modCount;
modCount = client.callProcedure("@AdHoc", "SELECT * FROM BLAH;").getResults()[0];
System.out.println("data inserted to table BLAH:\n" + modCount);
int rowct = modCount.getRowCount();
// Call validate partitioning to check if we are good.
VoltTable valTable;
valTable = client.callProcedure("@ValidatePartitioning", null, null).getResults()[0];
System.out.println("Validate for BLAH:\n" + valTable);
while (valTable.advanceRow()) {
long miscnt = valTable.getLong("MISPARTITIONED_ROWS");
assertEquals(miscnt, 0);
}
BufferedReader csvreport = new BufferedReader(new FileReader(CSVLoader.pathReportfile));
int lineCount = 0;
String line;
String promptMsg = "Number of rows successfully inserted:";
String promptFailMsg = "Number of rows that could not be inserted:";
int invalidlinecnt = 0;
while ((line = csvreport.readLine()) != null) {
if (line.startsWith(promptMsg)) {
String num = line.substring(promptMsg.length());
lineCount = Integer.parseInt(num.replaceAll("\\s", ""));
}
if (line.startsWith(promptFailMsg)) {
String num = line.substring(promptFailMsg.length());
invalidlinecnt = Integer.parseInt(num.replaceAll("\\s", ""));
}
}
csvreport.close();
System.out.println(String.format("The rows infected: (%d,%s)", lineCount, rowct));
assertEquals(lineCount, rowct);
// assert validLineCnt specified equals the successfully inserted lineCount
assertEquals(validLineCnt, lineCount);
assertEquals(invalidLineCnt, invalidlinecnt);
}
public static void main(String[] args) {
final boolean precomputed = false;
Instances trainData = null, testData = null;
final boolean crossValidation = args.length == 1;
try {
String dataFile = args[0];
System.err.println("INFO: Loading dataset from '" + dataFile + "' ...");
CSVLoader csvLoader = new CSVLoader();
// csvLoader.setStringAttributes("first"); // id
// csvLoader.setNominalAttributes("last"); // label
csvLoader.setNominalAttributes("first,last"); // id, label
csvLoader.setSource(new FileInputStream(dataFile));
Instances data = csvLoader.getDataSet();
if (!crossValidation) {
System.err.println(
"INFO: Loading splits from '" + args[1] + "' (train), '" + args[2] + "' (test) ...");
Set<String> trainLabels =
Collections.unmodifiableSet(new HashSet<String>(readAsList(args[1])));
Set<String> testLabels =
Collections.unmodifiableSet(new HashSet<String>(readAsList(args[2])));
{
Set<String> intersection = new HashSet<String>(trainLabels);
intersection.retainAll(testLabels);
if (!intersection.isEmpty())
throw new IllegalStateException(
"Train and test sets intersect: " + Arrays.toString(intersection.toArray()));
}
RemoveWithStringValues rwsv = new RemoveWithStringValues();
rwsv.setAttributeIndex("first");
trainData = rwsv.setValues(trainLabels).split(data, true, false);
testData = rwsv.setValues(testLabels).split(data, true, false);
if (trainData.classIndex() == -1) trainData.setClassIndex(trainData.numAttributes() - 1);
if (testData.classIndex() == -1) testData.setClassIndex(testData.numAttributes() - 1);
data = null;
} else if (!precomputed) {
if (data.classIndex() == -1) data.setClassIndex(data.numAttributes() - 1);
data.deleteWithMissingClass();
}
// feature selection
// if (crossValidation)
// {
// InfoGainFeatureSelection igfs = new InfoGainFeatureSelection(10,
// true);
// igfs.build(data);
//
// System.err.println("INFO: Selected " +
// igfs.selectedAttributes().size() +" of " +
// (data.numAttributes()-2) + " attributes.");
//
// Remove r = new Remove();
// r.setAttributeIndices("first," + igfs.selectedAttributeRange()
// +",last");
// r.setInvertSelection(true);
//
// r.setInputFormat(data);
// data = Filter.useFilter(data, r);
// if (data.classIndex() == -1)
// data.setClassIndex(data.numAttributes() - 1);
// }
// standardize
// {
// Standardize st = new Standardize();
//
// st.setInputFormat(data);
// data = Filter.useFilter(data, st);
// if (data.classIndex() == -1)
// data.setClassIndex(data.numAttributes() - 1);
// }
// weighting to balance classes
// {
// System.err.println("INFO: Weighting instances...");
// // collect statistics
// int[] classSupport = new int[data.numClasses()];
// int hasClass = 0;
// for (int i = 0; i < data.numInstances(); i++) {
// if (data.instance(i).classIsMissing())
// continue;
// classSupport[(int) data.instance(i).classValue()]++;
// hasClass++;
// }
//
// // calculate weights
// double[] classWeight = new double[data.numClasses()];
// final int smoothFactor = 2;
// double expectedFrequency = 1.0d / data.numClasses();
// for (int i = 0; i < data.numClasses(); i++) {
// final double frequency = classSupport[i]
// / (double) hasClass;
// final double ratio = expectedFrequency / frequency;
// classWeight[i] = (smoothFactor + ratio)/(smoothFactor + 1);
// System.err.println("INFO: Class '" +
// data.classAttribute().value(i) + "' instance weight set to " +
// (float)classWeight[i] );
// }
//
// for (int i = 0; i < data.numInstances(); i++) {
// if (data.instance(i).classIsMissing())
// continue;
// Instance inst = data.instance(i);
// inst.setWeight(classWeight[(int) inst.classValue()]);
// }
//
// }
if (crossValidation) {
for (int i = 0; i < data.numAttributes(); i++) {
if (i == 0 || i == data.classIndex()) continue;
Attribute a = data.attribute(i);
if (!a.isNumeric()) throw new IllegalStateException("attribute is not numeric: " + a);
}
}
final Classifier c;
{
if (!precomputed) {
// classifier
SMO smo = new SMO();
// c = smo;
FilteredClassifier smoRegex = prefixFilteredClassifier(new SMO(), "regex_");
HackedClassifier smoRegexHacked = new HackedClassifier();
smoRegexHacked.setSureClasses(Collections.singleton(Label.DM_Dermatology.code));
smoRegexHacked.setClassifier(smoRegex);
PrecomputedClassifier pc =
new PrecomputedClassifier(
Maps.readStringStringMapFromFile(
new File("resources/classificationResults_sampleRun.train+test.csv"),
Charset.forName("US-ASCII"),
",",
new HashMap<String, String>(),
true,
1));
pc.setSmoothFactor(0.7);
weka.classifiers.meta.Vote v = new Vote();
v.setOptions(new String[] {"-R", "AVG"});
v.setClassifiers(
new Classifier[] {wrapRemoveFirst(smo), wrapRemoveFirst(smoRegexHacked), pc});
// c = v;
c = wrapRemoveFirst(smo);
// GaussianProcesses gp = new GaussianProcesses ();
// c = gp;
// LADTree t = new LADTree();
// c = t;
// LibSVM svm = new LibSVM();
// c = svm;
// JRip jrip = new JRip();
// c = jrip;
// LogitBoost lb = new LogitBoost();
// lb.setClassifier(new SMOreg());
// lb.setDebug(true);
// c = lb;
// BayesNet bn = new BayesNet();
// // c = bn;
//
// J48 j48 = new J48();
// // j48.setMinNumObj(10);
// // c = j48;
//
// ClassificationViaRegression cvr = new
// ClassificationViaRegression();
// cvr.setClassifier(new SMOreg());
// c = cvr;
// Vote v = new Vote();
// v.setClassifiers(new Classifier[] {smo, bn, j48});
// c = v;
// // HIERARCHY -- groups
// HierachyNode root = new HierachyNode("ROOT");
// Map<Group, HierachyNode> groupNodes = new HashMap<Group,
// HierachyNode>();
// for (Group g : Group.values()) {
// HierachyNode groupNode = new HierachyNode(g.name());
// root.addChild(groupNode);
// groupNodes.put(g, groupNode);
// }
// for (Label l : Label.values()) {
// if (l.code != null) {
// HierachyNode labelNode = new HierachyNode(l.code);
// groupNodes.get(l.group).addChild(labelNode);
// } else
// System.err
// .println("WARNING: Skipping label without code "
// + l);
// }
// System.err.println(root.toString());
//
// HierarchicalClassifier hc = new
// HierarchicalClassifier(smo,
// root);
// c = hc;
// // HIERARCHY -- custom
// HierachyNode root = new HierachyNode("ROOT");
// HierachyNode radiology = new HierachyNode("RADIOLOGY");
// HierachyNode graphic = new HierachyNode("GRAPHIC");
// HierachyNode photo = new HierachyNode("PHOTO");
// root.addChild(radiology);
// root.addChild(graphic);
// root.addChild(photo);
// for (Label l : Label.values()) {
// if (l.code != null) {
// HierachyNode labelNode = new HierachyNode(l.code);
//
// switch (l) {
// case _3D_ThreeDee:
// graphic.addChild(labelNode);
// break;
// case AN_Angiography:
// radiology.addChild(labelNode);
// break;
// case CM_CompoundFigure:
// graphic.addChild(labelNode);
// break;
// case CT_ComputedTomography:
// radiology.addChild(labelNode);
// break;
// case DM_Dermatology:
// graphic.addChild(labelNode);
// break;
// case DR_Drawing:
// graphic.addChild(labelNode);
// break;
// case EM_ElectronMicroscopy:
// photo.addChild(labelNode);
// break;
// case EN_Endoscope:
// photo.addChild(labelNode);
// break;
// case FL_Fluorescense:
// radiology.addChild(labelNode);
// break;
// case GL_Gel:
// graphic.addChild(labelNode);
// break;
// case GR_GrossPathology:
// photo.addChild(labelNode);
// break;
// case GX_Graphs:
// graphic.addChild(labelNode);
// break;
// case HX_Histopathology:
// radiology.addChild(labelNode);
// break;
// case MR_MagneticResonance:
// radiology.addChild(labelNode);
// break;
// case PX_Photo:
// photo.addChild(labelNode);
// break;
// case RN_Retinograph:
// radiology.addChild(labelNode);
// break;
// case US_Ultrasound:
// radiology.addChild(labelNode);
// break;
// case XR_XRay:
// radiology.addChild(labelNode);
// break;
// default:
// throw new IllegalStateException(l.toString());
// }
//
// } else
// System.err
// .println("WARNING: Skipping label without code "+ l);
// }
// System.err.println(root.toString());
//
// HierarchicalClassifier hc = new
// HierarchicalClassifier(smo,
// root);
// c = hc;
// Cluster membership
// EM em = new EM();
// em.setNumClusters(Label.values().length); // TODO
// heuristics
//
// ClusterMembership cm = new AddClusterMembership();
// cm.setDensityBasedClusterer(em);
// cm.setIgnoredAttributeIndices(String.valueOf(data.classIndex()+1));
// // ignore class label
//
// FilteredClassifier fc = new FilteredClassifier();
// fc.setClassifier(hc);
// fc.setFilter(cm);
//
// c = fc;
// ASEvaluation ae = new InfoGainAttributeEval();
//
// Ranker ranker = new Ranker();
// ranker.setNumToSelect(data.numAttributes()/2);
// AttributeSelectedClassifier asc = new
// AttributeSelectedClassifier();
// asc.setClassifier(smo);
// asc.setEvaluator(ae);
// asc.setSearch(ranker);
//
// c = asc;
} else {
c =
new PrecomputedClassifier(
Maps.readStringStringMapFromFile(
new File("resources/classificationResults_sampleRun.txt"),
Charset.forName("US-ASCII"),
" ",
new HashMap<String, String>(),
true));
}
}
c.setDebug(true);
if (precomputed) evaluateClassifier(c, data, 10);
else if (crossValidation) evaluateClassifier(c, data, 10);
else
// split
evaluateClassifier(c, trainData, testData);
} catch (Exception e) {
System.err.println("ERROR: " + e.getMessage());
e.printStackTrace();
System.exit(1);
}
}
