Inspect a native classification tree

// Iris: Fisher, R. (1936). UCI. https://doi.org/10.24432/C56C76
// CC BY 4.0: https://creativecommons.org/licenses/by/4.0/
// Original UCI records retained; class names mapped to zero-based IDs.
// Full attribution and variant notes: datasets/iris/README.md.
import aner.dataset;
import aner.tensor;
import aner.ml;
import aner.metrics;

fn main() -> Unit {
    let flowers = dataset.iris();
    let partitions = dataset.split(flowers, 0.2, 2026);
    let training = dataset.train(partitions);
    let testing = dataset.test(partitions);

    // Choose these controls before inspecting the held-out test results.
    // Numeric binary splits support multiple classes. Scaling is unnecessary.
    let model: DecisionTreeClassifier = ml.fit_tree(
        features: dataset.features(training),
        labels: dataset.targets(training),
        criterion: "gini",
        max_depth: 4,
        min_samples_split: 4,
        min_samples_leaf: 2);
    let predicted = model.predict(dataset.features(testing));
    print("Dataset citation"); print(dataset.citation(flowers));
    print("Dataset SHA256"); print(dataset.sha256(flowers));
    print("Split seed"); print(2026);
    print("Training rows"); print(dataset.rows(training));
    print("Held-out rows"); print(dataset.rows(testing));
    print("Held-out accuracy"); print(metrics.accuracy(predicted, dataset.targets(testing)));
    print("Tree depth"); print(model.depth());
    print("Tree nodes"); print(model.node_count());
    print("Feature indices: 0 sepal length, 1 sepal width, 2 petal length, 3 petal width (cm)");
    print(model.tree_text());

    let probabilities = model.predict_proba(dataset.features(testing));
    let class_ids = model.classes();
    print("First held-out row: class ID then leaf class frequency (not calibrated)");
    var class_index = 0;
    while class_index < tensor.rows(class_ids) {
        print(tensor.value(class_ids, class_index, 0));
        print(tensor.value(probabilities, 0, class_index));
        class_index = class_index + 1;
    }
    // Files are created in the current working directory; existing paths are refused.
    // Choose new names or remove earlier exports before running this example again.
    // SVG opens directly in a browser and needs no Graphviz installation.
    model.export_tree("iris-decision-tree.svg");
    model.export_tree("iris-decision-tree.dot");
    model.export_tree("iris-decision-tree.txt");
    print("Exported iris-decision-tree.svg, .dot and .txt");
}