Refactoring.

Author: felipexw

src/main/java/com/github/felipexw/classifiers/bayes/MultinomialNaiveBayesClassifier.java

@@ -54,7 +54,7 @@ public void calculateProbs(List<LabeledInstance> instanceList) {
         if (instanceList == null || instanceList.isEmpty())
             throw new IllegalArgumentException("Args can't be null");
 
-        for (LabeledInstance<Label, Model> instance : instanceList) {
+        for (LabeledInstance<Model> instance : instanceList) {
             String key = instance.getLabel().toString();
             if (!labels.containsKey(key)) {
                 labels.put(instance.getLabel().toString(), 1);
@@ -86,8 +86,8 @@ public void calculatePosterioriProbability(LabeledInstance instance) {
          */
     }
 
-    private void countFromLabels(List<LabeledInstance<Label, Model>> instances, LabeledInstance<Label, Model> instance) {
-        for (LabeledInstance<Label, Model> featuresInstance : instances) {
+    private void countFromLabels(List<LabeledInstance<Model>> instances, LabeledInstance<Model> instance) {
+        for (LabeledInstance<Model> featuresInstance : instances) {
             if (featuresInstance.getLabel().toString().equalsIgnoreCase(instance.getLabel().toString())) {
                 featuresInstance.setCount(featuresInstance.getCount() + 1);
             } else {

src/main/java/com/github/felipexw/classifiers/neighbors/SimpleKNNClassifier.java

@@ -142,8 +142,11 @@ protected List<Neighbor> getNeighborsWithDistanceFromARootNeighboor(Neighbor nei
         LabeledInstance instance = neighbor.getInstance();
 
         for (int j = -1; j < instances.size() - 1; j++) {
-            LabeledInstance neighborInstance = instances.get(j + 1);
-            double similarity = similarityCalculator.calculate(instance.getFeatures(), neighborInstance.getFeatures());
+            LabeledInstance<Model> neighborInstance = instances.get(j + 1);
+            Model<Double> model = instance.getModel();
+            Model<Double> neighborModel = neighborInstance.getModel();
+
+            double similarity = similarityCalculator.calculate(model.getData(), neighborModel.getData());
             Neighbor neighborRoot = new Neighbor(neighborInstance, similarity, new DoubleFeatureExtractor());
             neighbors.add(neighborRoot);
 
@@ -193,7 +196,7 @@ protected List<Neighbor> getAllNeighbors(Instance labeledInstance) {
         List<Neighbor> neighborses = new ArrayList<>();
         for (short i = 0; i < instances.size(); i++) {
             LabeledInstance trainingInstance = instances.get(i);
-            double distance = similarityCalculator.calculate(labeledInstance.getFeatures(), trainingInstance.getFeatures());
+            double distance = similarityCalculator.calculate(labeledInstance.getModel().getData(), trainingInstance.getModel().getData());
 
             Neighbor neighbor = new Neighbor(trainingInstance, distance, featureExtractor);
             neighborses.add(neighbor);
@@ -207,7 +210,7 @@ public Prediction vote(List<Neighbor> neighbors) {
         Map<String, Integer> votes = new HashMap<>();
 
         for (Neighbor neighbor : neighbors) {
-            LabeledInstance<Label, Model> instance = neighbor.getInstance();
+            LabeledInstance<Model> instance = neighbor.getInstance();
             if (!votes.containsKey(instance.getLabel()))
                 votes.put(instance.getLabel().toString(), 1);
 
@@ -222,7 +225,6 @@ public Prediction vote(List<Neighbor> neighbors) {
         int nearestNeighborIndex = getIndexOfNearestNeighboorVoted(mostVotedLabel, neighbors);
         Neighbor neighbor = neighbors.get(nearestNeighborIndex);
 
-
         return new Prediction(mostVotedLabel, neighbor.getDistance() / 100);
     }
 
@@ -248,7 +250,7 @@ protected int getIndexOfNearestNeighboorVoted(String label, List<Neighbor> neigh
 
         for (int i = 0; i < neighbors.size(); i++) {
             Neighbor neighbor = neighbors.get(i);
-            LabeledInstance<Label, Model> instance = neighbor.getInstance();
+            LabeledInstance<Model> instance = neighbor.getInstance();
             if (instance.getLabel().toString().equalsIgnoreCase(label) && neighbor.getDistance() < distance) {
                 distance = neighbor.getDistance();
                 index = i;

src/main/java/com/github/felipexw/core/Instance.java

@@ -5,8 +5,14 @@
 /**
  * Created by felipe.appio on 23/08/2016.
  */
-public abstract class Instance<F> {
-    protected  List<F> features;
+public abstract class Instance<F extends Model> {
+    protected final F model;
 
-    public abstract  List<F> getFeatures();
+    public Instance(F model) {
+        this.model = model;
+    }
+
+    public Model getModel() {
+        return model;
+    }
 }

src/main/java/com/github/felipexw/core/Label.java

@@ -6,28 +6,24 @@
 /**
  * Created by felipe.appio on 23/08/2016.
  */
-public class LabeledInstance<L extends Label, F extends Model> extends Instance<F> {
-    protected L label;
+public class LabeledInstance<F extends Model> extends Instance<F> {
+    protected String label;
     private int count;
 
-    public LabeledInstance(List<F> features,L label) {
+    public LabeledInstance(String label, F model) {
+        super(model);
         this.label = label;
-        this.features = features;
     }
 
-    public L getLabel() {
+    public String getLabel() {
         return this.label;
     }
 
+
     public void setCount(int count) {
         this.count = count;
     }
 
-    @Override
-    public List<F> getFeatures() {
-
-        return null;
-    }
 
     public int getCount() {
         return count;
@@ -38,7 +34,7 @@ public boolean equals(Object o) {
         if (this == o) return true;
         if (o == null || getClass() != o.getClass()) return false;
 
-        LabeledInstance<?, ?> that = (LabeledInstance<?, ?>) o;
+        LabeledInstance<?> that = (LabeledInstance<?>) o;
 
         return label.equals(that.label);
 

src/main/java/com/github/felipexw/core/LabeledInstance.java

@@ -8,18 +8,20 @@
 /**
  * Created by felipe.appio on 31/08/2016.
  */
-public class Model implements FeatureVector {
+public class Model<T> implements FeatureVector {
 
     protected FeatureExtractor featureExtractor;
-    protected List features;
+    protected List<T> features;
 
     public Model(FeatureExtractor featureExtractor) {
         this.featureExtractor = featureExtractor;
     }
 
     @Override
-    public List getData() {
-        return  featureExtractor.extract(features);
+    public List<T> getData() {
+        if (features == null)
+            features = featureExtractor.extract(features);
+        return features;
     }
 
 }

src/main/java/com/github/felipexw/core/Model.java

@@ -3,16 +3,16 @@
 /**
  * Created by felipe.appio on 24/08/2016.
  */
-public class Prediction<L> {
-    private final L label;
+public class Prediction{
+    private final String label;
     private final double score;
     private int count;
 
-    public L getLabel() {
+    public String getLabel() {
         return label;
     }
 
-    public Prediction(L label, double score) {
+    public Prediction(String label, double score) {
         this.label = label;
         this.score = score;
     }
@@ -29,21 +29,6 @@ public double getScore() {
         return score;
     }
 
-    @Override
-    public boolean equals(Object o) {
-        if (this == o) return true;
-        if (!(o instanceof Prediction)) return false;
-
-        Prediction that = (Prediction) o;
-
-        return label.equals(that.label);
-    }
-
-    @Override
-    public int hashCode() {
-        return label.hashCode();
-    }
-
     @Override
     public String toString() {
         return "Prediction{" +

src/main/java/com/github/felipexw/core/Prediction.java

@@ -1,20 +1,14 @@
-import com.github.felipexw.classifiers.neighbors.KNNClassifier;
-import com.github.felipexw.classifiers.neighbors.Neighbor;
 import com.github.felipexw.classifiers.neighbors.SimpleKNNClassifier;
-import com.github.felipexw.core.Label;
 import com.github.felipexw.core.Model;
 import com.github.felipexw.core.Prediction;
 import com.github.felipexw.core.extraction.DoubleFeatureExtractor;
 import com.github.felipexw.core.extraction.FeatureExtractor;
 import com.github.felipexw.evaluations.metrics.EuclidianSimilarityCalculator;
 import com.github.felipexw.core.LabeledInstance;
 import com.google.common.truth.Truth;
-import javafx.scene.control.Labeled;
 import org.junit.Before;
 import org.junit.Test;
-import com.google.common.truth.Truth.*;
 
-import java.sql.Array;
 import java.util.*;
 
 /**
@@ -24,9 +18,9 @@
 */
 public class SimpleKNNClassifierTest {
     static class TestModel extends Model{
-        private List<String> stringFeatures;
+        private List<Double> stringFeatures;
 
-        public TestModel(FeatureExtractor featureExtractor, List<String> stringFeatures) {
+        public TestModel(FeatureExtractor featureExtractor, List<Double> stringFeatures) {
             super(featureExtractor);
             this.features = stringFeatures;
         }
@@ -60,31 +54,31 @@ public void it_should_predict_a_negative_label(){
         the algorithm must predict the label (which its positive or negative) for the point E(1,3)
         */
 
-        Label positiveLabel = new Label("positive");
-        Label negativeLabel = new Label("negative");
+        String positiveLabel = new String("positive");
+        String negativeLabel = new String("negative");
         FeatureExtractor featureExtractor =   new DoubleFeatureExtractor();
 
-        TestModel t1 = new TestModel(featureExtractor, Arrays.asList("3.0", "4.0"));
-        LabeledInstance<Label, Model> pointA = new LabeledInstance<Label, Model>(Arrays.asList(t1), negativeLabel);
+        TestModel t1 = new TestModel(featureExtractor, Arrays.asList(3d, 4d));
+        LabeledInstance<Model> pointA = new LabeledInstance<Model>(negativeLabel, t1);
 
-        TestModel t2 = new TestModel(featureExtractor, Arrays.asList("3.0", "2.0"));
-        LabeledInstance<Label, Model> pointB = new LabeledInstance<Label, Model>(Arrays.asList(t2), negativeLabel);
+        TestModel t2 = new TestModel(featureExtractor, Arrays.asList(3d, 2d));
+        LabeledInstance<Model> pointB = new LabeledInstance<Model>(negativeLabel, t2);
 
-        TestModel t3 = new TestModel(featureExtractor, Arrays.asList("4.0", "1.0"));
-        LabeledInstance<Label, Model> pointC = new LabeledInstance<Label, Model>(Arrays.asList(t3), negativeLabel);
+        TestModel t3 = new TestModel(featureExtractor, Arrays.asList(4d, 1d));
+        LabeledInstance<Model> pointC = new LabeledInstance<Model>(positiveLabel, t3);
 
-        TestModel t4 = new TestModel(featureExtractor, Arrays.asList("5.0", "5.0"));
-        LabeledInstance<Label, Model> pointD = new LabeledInstance<Label, Model>(Arrays.asList(t2), negativeLabel);
+        TestModel t4 = new TestModel(featureExtractor, Arrays.asList(50d, 5d));
+        LabeledInstance<Model> pointD = new LabeledInstance<Model>(positiveLabel, t4);
 
-        TestModel predictingTest = new TestModel(featureExtractor, Arrays.asList("1.0", "3.0"));
-        LabeledInstance<Label, Model> pointE = new LabeledInstance<Label, Model>(Arrays.asList(predictingTest), negativeLabel);
+        TestModel predictingTest = new TestModel(featureExtractor, Arrays.asList(1d, 3d));
+        LabeledInstance<Model> pointE = new LabeledInstance<Model>(negativeLabel,predictingTest);
 
         classifier.setK(5);
         classifier.train(Arrays.asList(pointA, pointB, pointC, pointD));
         Prediction predictedInstance = classifier.predict(pointE);
 
         Truth.assertThat(predictedInstance.getLabel())
-                .isEqualTo(negativeLabel);
+                .isEqualTo(positiveLabel);
     }
 
 }