Cardiolog is a medical decision-support system designed to address the lack of interpretability and accountability in clinical AI.
Instead of returning opaque risk scores, the system provides explicit, human-readable justifications for each prediction using a neuro-symbolic architecture.

The project explores how statistical learning can be translated into logical, inspectable reasoning pipelines, making AI outputs more suitable for high-stakes and regulated environments.

• 🔍 Glass-Box Decision Support
  Translates opaque Random Forest decision paths into transparent logical rules that can be inspected, audited, and debated.

• 🗣️ Explainable Justifications
  Generates structured, natural-language explanations such as:
  “High risk BECAUSE cholesterol > 240 AND age > 50”, enabling clinician-level understanding of model behavior.

• 🧠 Hybrid Intelligence Architecture
  Combines the predictive strengths of Machine Learning with the reasoning and traceability of symbolic logic (Prolog / s(CASP)).

• 🧬 Dynamic Patient Modeling
  New patient data is automatically converted into logical facts, enabling real-time inference and explanation without retraining the model.

The system is organized as a three-layer pipeline that converts statistical patterns into logical proofs:

• Trains a Random Forest Classifier on the Cleveland Heart Disease dataset.
• Captures predictive structure while accepting limited interpretability at this stage.

• A custom Tree-to-Logic compiler (translator.py) traverses each decision tree and transpiles its paths into s(CASP)-compatible predicates.
• Each tree is treated as an independent reasoning agent, contributing partial evidence.

• Aggregates logical rules and patient-specific facts.
• Executes queries using s(CASP) to compute stable models.
• Produces a structured justification tree, rendered as HTML for inspection.

/
├── cardiolog/           # Core source code
│   ├── translator.py    # Decision Tree → Logic compiler
│   ├── main.py          # Pipeline orchestration and inference
│   └── ...
├── data/                # Clinical datasets (CSV)
├── prolog/              # Generated logical knowledge base
├── templates/           # HTML explanation templates
└── output/              # Reasoning traces and justification trees