Skip to content

GA-Optimized Decision Trees

🌳 GA-Optimized Decision Trees

Evolving decision trees that balance accuracy and interpretability using multi-objective genetic algorithms.

Achieve 46–82% smaller trees with statistically equivalent accuracy — validated with 20-fold cross-validation (p > 0.05 on all benchmarks).

82%
Max size reduction
p > 0.05
Accuracy parity
3
Benchmark datasets
20‑CV
Fold validation

🚀 Quick Start

```bash
git clone https://github.com/ibrah5em/ga-optimized-trees.git
cd ga-optimized-trees
python -m venv venv && source venv/bin/activate
pip install -e .          # core only
pip install -e .[all]     # all features
```

```bash
python scripts/train.py --config configs/paper.yaml --dataset iris
```

```python
from ga_trees import GAEngine, GAConfig, TreeInitializer, FitnessCalculator, Mutation
from ga_trees.data import DatasetLoader
from ga_trees.fitness import TreePredictor
import numpy as np

data = DatasetLoader().load_dataset("iris", test_size=0.2)
X_train, y_train = data["X_train"], data["y_train"]

config   = GAConfig(population_size=80, n_generations=40)
init     = TreeInitializer(X_train.shape[1], len(np.unique(y_train)), max_depth=6)
fitness  = FitnessCalculator(accuracy_weight=0.68, interpretability_weight=0.32)
mutation = Mutation(X_train.shape[1], {i: (X_train[:,i].min(), X_train[:,i].max())
                                        for i in range(X_train.shape[1])})

engine = GAEngine(config, init, fitness.calculate_fitness, mutation)
best   = engine.evolve(X_train, y_train, verbose=True)
print(f"Tree: {best.get_num_nodes()} nodes, depth {best.get_depth()}")
```

```bash
python scripts/experiment.py --config configs/paper.yaml
```

📚 Documentation

⚙️
Installation
All install methods, extras, and platform notes

⚡
Quick Start
Up and running in 5 minutes

🏗️
Architecture
System design and component overview

🧬
Genetic Algorithm
Selection, crossover, mutation internals

📖
API Reference
Full class and method documentation

📊
Results
Benchmark tables and statistical tests

🌸
Examples
Iris, medical, and credit scoring walkthroughs

❓
FAQ
Common questions and troubleshooting

📈 Benchmark Results

Dataset GA Accuracy CART Accuracy GA Nodes CART Nodes Size Reduction
Iris 94.55 ± 8.07% 92.41 ± 10.43% 7.4 16.4 55%
Wine 88.19 ± 10.39% 87.22 ± 10.70% 10.7 20.7 48%
Breast Cancer 91.05 ± 5.60% 91.57 ± 3.92% 6.5 35.5 82%

All p-values > 0.05 — accuracy difference is not statistically significant.

🆚 How It Compares

Aspect CART Random Forest GA-Optimized
Optimization Greedy (local) Ensemble Global (evolutionary)
Objectives Accuracy only Accuracy only Multi-objective
Interpretability No control Black box Explicit control ✓
Tree size Often large N/A Controllable ✓

v1.0.0 Python 3.8+ MIT License