Add Auto-Weka to pipeline

Author: oshando

swan-pipeline/src/main/java/de/fraunhofer/iem/swan/features/FeatureSetSelector.java

@@ -15,6 +15,8 @@ public IFeatureSet select(Dataset dataset, SwanOptions options) {
                 mekaFeatureSet.createFeatures();
                 return mekaFeatureSet;
             case WEKA:
+            case MLPLAN:
+            case AUTOWEKA:
                 WekaFeatureSet wekaFeatureSet = new WekaFeatureSet(dataset, options);
                 wekaFeatureSet.createFeatures();
                 return wekaFeatureSet;

swan-pipeline/src/main/java/de/fraunhofer/iem/swan/model/ModelEvaluator.java

@@ -7,6 +7,7 @@
 import de.fraunhofer.iem.swan.features.WekaFeatureSet;
 import de.fraunhofer.iem.swan.io.dataset.SrmList;
 import de.fraunhofer.iem.swan.io.dataset.SrmListUtils;
+import de.fraunhofer.iem.swan.model.toolkit.AutoWeka;
 import de.fraunhofer.iem.swan.model.toolkit.MLPlan;
 import de.fraunhofer.iem.swan.model.toolkit.Meka;
 import de.fraunhofer.iem.swan.model.toolkit.Weka;
@@ -28,7 +29,8 @@ public class ModelEvaluator {
     public enum Toolkit {
         WEKA,
         MEKA,
-        MLPLAN
+        MLPLAN,
+        AUTOWEKA
     }
 
     public enum Phase {
@@ -64,14 +66,17 @@ public void trainModel() {
                 processResults(meka.trainModel());
                 break;
             case WEKA:
-                logger.info("Evaluating model with WEKA");
-                Weka weka = new Weka((WekaFeatureSet) features, options);
-                weka.trainModel();
+                Weka weka = new Weka((WekaFeatureSet) features, options, methods);
+                processResults(weka.trainModel());
+                break;
+            case AUTOWEKA:
+                AutoWeka autoWeka = new AutoWeka((WekaFeatureSet) features, options);
+                autoWeka.trainModel();
                 break;
             case MLPLAN:
                 logger.info("Evaluating model with ML-PLAN");
-                MLPlan mlPlan = new MLPlan();
-                mlPlan.evaluateDataset(((WekaFeatureSet) features).getInstances().get("train"));
+                MLPlan mlPlan = new MLPlan((WekaFeatureSet) features, options);
+                mlPlan.trainModel();
                 break;
         }
     }

swan-pipeline/src/main/java/de/fraunhofer/iem/swan/model/toolkit/AutoWeka.java

@@ -0,0 +1,83 @@
+package de.fraunhofer.iem.swan.model.toolkit;
+
+import de.fraunhofer.iem.swan.cli.SwanOptions;
+import de.fraunhofer.iem.swan.features.WekaFeatureSet;
+import de.fraunhofer.iem.swan.model.ModelEvaluator;
+import de.fraunhofer.iem.swan.util.Util;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+import weka.classifiers.Evaluation;
+import weka.classifiers.meta.AutoWEKAClassifier;
+import weka.core.Instances;
+
+import java.util.HashMap;
+import java.util.LinkedHashMap;
+
+
+public class AutoWeka {
+
+    private WekaFeatureSet features;
+    private SwanOptions options;
+    private static final Logger logger = LoggerFactory.getLogger(AutoWeka.class);
+
+    public AutoWeka(WekaFeatureSet features, SwanOptions options) {
+        this.features = features;
+        this.options = options;
+    }
+
+    /**
+     * Trains and evaluates the model with the given training data and specified classification mode.
+     *
+     * @return Hashmap containing the name of the classifier and it's F-Measure
+     */
+    public HashMap<String, HashMap<String, String>> trainModel() {
+
+        switch (ModelEvaluator.Phase.valueOf(options.getPhase().toUpperCase())) {
+            case VALIDATE:
+
+                //Phase 1: classify SRM classes
+                for (String srm : options.getSrmClasses())
+                    runManualEvaluation(features.getTrainInstances().get(srm));
+
+                //Filter methods from CWE instances that were not classified into one of the SRM classes
+                //Phase 2: classify CWE classes
+                for (String cwe : options.getCweClasses())
+                    runManualEvaluation(features.getTrainInstances().get(cwe));
+
+                return null;
+            case PREDICT:
+
+        }
+        return null;
+    }
+
+    /**
+     * @return
+     */
+    public HashMap<String, HashMap<String, String>> runManualEvaluation(Instances instances) {
+
+        String category = instances.attribute(instances.numAttributes() - 1).name();
+        instances.setClass(instances.attribute(instances.numAttributes() - 1));
+        String instancesFile = Util.exportInstancesToArff(instances, category + "-autoweka");
+
+        LinkedHashMap<String, HashMap<String, String>> fMeasure = new LinkedHashMap<>();
+        logger.info("Selecting model for {} using Auto-Weka:  timelimit= {}, instances={}",
+                category, options.getTimeLimit(), instancesFile);
+
+        try {
+
+            String[] args = {"-t", instancesFile,
+                    "-seed", "1",
+                    "-no-cv",
+                    "-timeLimit", Integer.toString(options.getTimeLimit())};
+
+            String out = Evaluation.evaluateModel(new AutoWEKAClassifier(), args);
+            logger.info("Auto-Weka Results: {]}",out);
+
+            //autoWekaClassifier.buildClassifier(instances);
+        } catch (Exception e) {
+            throw new RuntimeException(e);
+        }
+        return fMeasure;
+    }
+}

swan-pipeline/src/main/java/de/fraunhofer/iem/swan/model/toolkit/Meka.java

@@ -35,7 +35,6 @@ public Meka(MekaFeatureSet features, SwanOptions options, Set<Method> methods) {
 
     /**
      * Trains and evaluates the model with the given training data and specified classification mode.
-     *
      */
     public SrmList trainModel() {