// TODO save dan load
@Override
public void writeHypothesis(OutputStream str) {
PrintStream printStream = new PrintStream(str);
printStream.println(k);
// save tipe dari classifier
printStream.println(classifier.getClass().getName());
printStream.flush();
classifier.writeHypothesis(str);
}
@Override
public void train(
int[] numInputCategory, int[][] inputCategory, int numOutputClass, int[] outputClass)
throws Exception {
shuffleTrainingSet(inputCategory, outputClass);
OfflineLearningNominalDataClassifier[] candidate = new OfflineLearningNominalDataClassifier[k];
double[] accuracy = new double[k];
int selected = 0;
for (int i = 0; i < k; i++) {
candidate[i] = classifier.copy();
// pisah set menjadi tr dan cv
ArrayList<int[]> trInputCategory = new ArrayList<int[]>();
ArrayList<Integer> trOutputClass = new ArrayList<Integer>();
ArrayList<int[]> cvInputCategory = new ArrayList<int[]>();
ArrayList<Integer> cvOutputClass = new ArrayList<Integer>();
for (int j = 0; j < i * inputCategory.length / k; j++) {
trInputCategory.add(inputCategory[j]);
trOutputClass.add(outputClass[j]);
}
for (int j = i * inputCategory.length / k; j < (i + 1) * inputCategory.length / k; j++) {
cvInputCategory.add(inputCategory[j]);
cvOutputClass.add(outputClass[j]);
}
for (int j = (i + 1) * inputCategory.length / k; j < inputCategory.length; j++) {
trInputCategory.add(inputCategory[j]);
trOutputClass.add(outputClass[j]);
}
candidate[i].train(
numInputCategory,
(int[][]) trInputCategory.toArray(new int[trInputCategory.size()][]),
numOutputClass,
trOutputClass.stream().mapToInt(Integer::intValue).toArray());
accuracy[i] =
calculateAccuracy(
candidate[i],
cvInputCategory.toArray(new int[cvInputCategory.size()][]),
cvOutputClass.stream().mapToInt(Integer::intValue).toArray());
if (accuracy[i] > accuracy[selected]) selected = i;
}
classifier = candidate[selected];
}
@Override
public void loadHypothesis(Scanner sc) {
k = sc.nextInt();
String className = sc.nextLine();
System.out.println(className);
Class cl;
try {
cl = Class.forName(className);
classifier = (OfflineLearningNominalDataClassifier) cl.newInstance();
} catch (ClassNotFoundException ex) {
Logger.getLogger(kFold.class.getName()).log(Level.SEVERE, null, ex);
} catch (InstantiationException ex) {
Logger.getLogger(kFold.class.getName()).log(Level.SEVERE, null, ex);
} catch (IllegalAccessException ex) {
Logger.getLogger(kFold.class.getName()).log(Level.SEVERE, null, ex);
}
classifier.loadHypothesis(sc);
}
@Override
public int predict(int[] inputCategory) throws Exception {
return classifier.predict(inputCategory);
}
public kFold(int k, OfflineLearningNominalDataClassifier classifier) {
this.k = k;
this.classifier = classifier.copy();
}
