Eldeewealth/Fairness-Aware-XAI-for-Responsible-Algorithmic-Trading: Fairness-Aware Explainable AI for Responsible Algorithmic Trading – v1.0.0

Fairness-Aware Explainable AI for Responsible Algorithmic Trading – v1.0.0 This release provides the archived research code accompanying the Springer Digital Finance submission: "Fairness-Aware Explainable AI for Responsible Algorithmic Trading" Overview This version contains the complete experimental pipeline for fairness-aware binary classification of next-day S&P 500 direction. The implementation integrates predictive modeling, fairness mitigation, explainability, calibration diagnostics, and trading simulation within a reproducible workflow. Included Components LightGBM baseline classifier Fairlearn Exponentiated Gradient fairness mitigation (Demographic Parity) Group-level fairness evaluation under volatility-regime stratification SHAP-based model explainability (baseline vs. fairness-mitigated) Calibration diagnostics (reliability curve, Brier score) Strategy-level backtesting framework MLflow experiment tracking Data acquisition and cleaning scripts Full end-to-end analysis notebook Reproducibility This release includes: datafile.py — S&P 500 data acquisition clean_sp500.py — cleaning and validation pipeline fairness_check.ipynb — complete experimental notebook requirements.txt — pinned core dependencies requirements-notebooks.txt — notebook environment extras The repository structure and environment specifications allow replication of the experiments described in the associated manuscript. Archival Purpose Version v1.0.0 is archived for citation, reproducibility, and long-term accessibility via Zenodo DOI minting.