Model Export and Loading

Save and load trained models.

Saving Models

import pickle
from pathlib import Path

# Prepare model data
model_data = {
    "tree": best_tree,
    "scaler": scaler,  # If used
    "feature_ranges": feature_ranges,
    "feature_names": feature_names,
    "class_names": class_names,
    "n_features": n_features,
    "n_classes": n_classes,
    "config": vars(ga_config),
    "metrics": {
        "train_accuracy": train_acc,
        "test_accuracy": test_acc,
        "test_f1": test_f1,
        "nodes": best_tree.get_num_nodes(),
        "depth": best_tree.get_depth(),
    },
}

# Save
output_path = Path("models/best_tree.pkl")
output_path.parent.mkdir(parents=True, exist_ok=True)
with open(output_path, "wb") as f:
    pickle.dump(model_data, f)

print(f"Model saved to: {output_path}")

Loading Models

import pickle

# Load
with open("models/best_tree.pkl", "rb") as f:
    model_data = pickle.load(f)

# Extract components
tree = model_data["tree"]
scaler = model_data.get("scaler")
feature_names = model_data.get("feature_names")
class_names = model_data.get("class_names")
metrics = model_data.get("metrics", {})

print(f"Loaded model:")
print(f"  Nodes: {metrics.get('nodes')}")
print(f"  Accuracy: {metrics.get('test_accuracy'):.4f}")

Making Predictions

from ga_trees.fitness.calculator import TreePredictor

# Prepare new data
X_new = load_new_data()
if scaler:
    X_new = scaler.transform(X_new)

# Predict
predictor = TreePredictor()
y_pred = predictor.predict(tree, X_new)

print(f"Predictions: {y_pred}")