Explainable GeoAI: Interpreting Socio-Spatial Patterns

An end-to-end spatial XAI pipeline combining XGBoost, SHAP, GeoShapley, and MGWR to uncover and visualize interpretable spatial effects in U.S. county-level voting data.

---

📂 Repository Structure

explainable-geoai/
├── data/
│   ├── raw/
│   │   ├── census/                   # raw ACS downloads
│   │   ├── shapefiles/              # county geometries
│   │   └── voting\_2021.csv          # raw vote share
│   └── processed/
│       ├── voting\_clean.csv         # cleaned tabular data
│       ├── voting\_features.csv      # with engineered features & spatial lags
│       ├── xgb\_automl\_model.pkl     # trained FLAML+XGBoost model
│       ├── shap\_explanations.csv    # SHAP outputs
│       ├── geoshapley\_explanations.csv # GeoShapley outputs
│       ├── mgwr\_coefficients.csv    # MGWR baseline
│       ├── bootstrap\_shap\_stats.csv # SHAP uncertainty stats
│       └── fairness\_metrics.csv     # spatial fairness gaps
├── src/
│   ├── data\_loader.py               # load & clean
│   ├── feature\_engineering.py       # spatial lags, GeoDataFrame
│   ├── model\_training.py            # FLAML + XGBoost training
│   ├── shap\_explainer.py            # Kernel SHAP wrapper
│   ├── geoshapley\_explainer.py      # GeoShapley computations
│   ├── mgwr\_comparison.py           # MGWR baseline scripts
│   ├── bootstrap\_uncertainty.py     # bootstrap SHAP stats
│   ├── spatial\_fairness.py          # compute residual‐fairness
│   └── config.py                    # paths & constants
├── dashboard/
│   └── app.py                       # Streamlit + Folium dashboard
├── docs/
│   ├── implementation\_notes.md      # detailed pipeline doc
│   └── paper\_summary.pdf            # summary of Li (2025) chapter
├── README.md                        # this file
└── requirements.txt                 # pip dependencies

⚙️ Installation

1. Clone repo

   git clone https://github.com/yourusername/explainable-geoai.git
   cd explainable-geoai

2. **Create & activate** a virtual environment

   ```bash
   python3 -m venv venv
   source venv/bin/activate   # macOS/Linux
   venv\Scripts\activate      # Windows

3. Install dependencies

   pip install -r requirements.txt

4. Download raw data

   • Place voting_2021.csv in data/raw/
   • Download ACS and shapefiles via src/download_census.py or manually.

🚀 Quick Start

1. Data & features

   python src/data_loader.py
   python src/feature_engineering.py

2. Train model

   python src/model_training.py

3. Generate explanations

   python src/shap_explainer.py
   python src/geoshapley_explainer.py
   python src/mgwr_comparison.py
   python src/bootstrap_uncertainty.py
   python src/spatial_fairness.py

4. Launch dashboard

   cd dashboard
   streamlit run app.py

📝 Scripts & Modules

• data_loader.py: cleans raw vote + ACS, saves voting_clean.csv.
• feature_engineering.py: builds spatial lags, exports voting_features.csv.
• model_training.py: uses FLAML to find best XGBoost; saves model.
• shap_explainer.py: Kernel SHAP over FLAML model → shap_explanations.csv.
• geoshapley_explainer.py: computes GeoShapley components → geoshapley_explanations.csv.
• mgwr_comparison.py: fits MGWR baseline → mgwr_coefficients.csv.
• bootstrap_uncertainty.py: bootstraps SHAP → bootstrap_shap_stats.csv.
• spatial_fairness.py: calculates fairness gaps → fairness_metrics.csv.
• dashboard/app.py: interactive Streamlit + Folium map.

📊 Dashboard Overview

• SHAP: county-level attributions, with uncertainty.
• GeoShapley: decomposed intrinsic (GEO), main, and interaction effects.
• MGWR/OLS: local regression coefficients for comparison.
• Fairness: residual differences across demographic groups.
• Download any CSV for offline analysis.

🧾 Citing

If you use this work, please cite:

Li, Ziqi (2025). Explainable AI in Spatial Analysis. In: Advances in Spatial Data Science, Springer.