Ankara EGO Bus & Rail System Analysis

Overview

A comprehensive spatial and temporal analysis of Ankara's EGO public transit system. The study examines service-demand mismatches, route efficiency, and structural patterns using grid-based spatial analysis, anomaly detection, and regime shift analysis.

Key Findings

• 173 days of bus data analyzed (Dec 2023 – Oct 2024)
• 315 days of metro/rail data (Dec 2023 – Nov 2024)
• 550 bus routes + 5 rail lines covering 1,382 grid cells (1km × 1km)
• 13 critical grids identified (both anomalous and unstable)
• 417 undersupply hotspots and 353 oversupply coldspots (LISA)
• Weekday vs Weekend: -28.8% passengers, -21% occupancy

Methodology

Data Pipeline

1. PDF Parsing → Extract daily bus & metro data from EGO PDF reports
2. Date Correction → Match reports to actual calendar dates
3. Grid Aggregation → 1km × 1km spatial grid, two distribution methods:
   • Connectivity-weighted: Passengers distributed by stop transfer power
   • Position-weighted: Terminal-biased distribution (30% first/last stops)

Analysis Methods

• Anomaly Detection: Autoencoder (PCA), Isolation Forest, LOF, Graph-based
• Regime Shift: Embedding stability, Daily clustering
• Spatial Analysis: Global/Local Moran's I (LISA)

Tech Stack

• Python 3.13 — pandas, numpy, scikit-learn

• Visualization — Matplotlib, Seaborn, Plotly, Folium

• ML — Scikit-learn (Isolation Forest, LOF, K-Means, PCA)

Quick Start

pip install pandas numpy matplotlib seaborn folium plotly scikit-learn pysal networkx pdfplumber

# Create grid data
python ai_analysis/scripts/create_daily_grid_data.py

# Run anomaly analysis
python ai_analysis/scripts/grid_anomaly_and_regime_analysis.py

# Visualize results
python ai_analysis/scripts/visualize_anomaly_regime_results.py

# Generate grid maps
python scripts/visualize_raw_features_grid.py --input data/daily_grid_data.csv

Data Sources

• EGO Genel Müdürlüğü — Ankara Metropolitan Municipality, Public Transit Authority
• Daily bus route reports & metro/rail passenger data (PDF)
• -Overpass and Google Geocoding API

License

Academic research use.

Citation

If you use this work, please cite:

@misc{kancan2026skeletaltrap,
  title        = {The Skeletal Trap: Mapping Spatial Inequality and Ghost Stops in Ankara's Transit Network},
  author       = {Kancan, Elifnaz},
  year         = {2026},
  eprint       = {2602.15470},
  archivePrefix= {arXiv},
  primaryClass = {physics.soc-ph},
  doi          = {10.48550/arXiv.2602.15470},
  url          = {https://arxiv.org/abs/2602.15470}
}