Machine Learning mit Java

@ApplicationScoped
public class DecisionTreeRepository {
    private DecisionTree model;

    private Map<String, Integer> speciesMap = new HashMap<>();

    public void setSpeciesMap(Map<String, Integer> map) {
        this.speciesMap = map;
    }

    public String getSpeciesLabel(int code) {
        // Reverse lookup (inefficient for large maps, but fine here)
        return speciesMap.entrySet()
                .stream()
                .filter(entry -> entry.getValue() == code)
                .map(Map.Entry::getKey)
                .findFirst()
                .orElse("Unknown");
    }

    public void train(List<Iris> irisList) {
        Formula formula = Formula.lhs("species");
        DataFrame data = DataFrame.of(Iris.class, irisList);

        // Hyperparameters can be set here
        DecisionTree.Options options = new DecisionTree.Options();

        model = DecisionTree.fit(formula, data, options);
    }

    public int[] predict(DataFrame input) {
        if (model == null) {
            throw new IllegalStateException("Model is not trained yet.");
        }
        return model.predict(input);
    }
}

//Austauschen
private DecisionTree model; // von
private RandomForest model; // durch

    private MultiLayerConfiguration getModelConfiguration() {
        return new NeuralNetConfiguration.Builder()
                .seed(123)
                .l2(0.0005)
                .weightInit(WeightInit.XAVIER)
                .activation(Activation.RELU)
                .updater(new Nesterovs(0.02, 0.9))
                .list()
                .layer(new DenseLayer.Builder().nIn(4).nOut(10).build())
                .layer(new OutputLayer.Builder(LossFunctions.LossFunction.MCXENT)
                        .nIn(10).nOut(3)  // 3 output classes: Setosa, Versicolor, Virginica
                        .activation(Activation.SOFTMAX)
                        .build())
                .setInputType(InputType.feedForward(4))  // 4 input features
                .build();
    }