AccessFlow Toronto — Pedestrian Impact Analysis

Machine learning analysis of Toronto road-restriction data and its impact on pedestrian network accessibility.

Quick Start

pip install -r requirements.txt
python generate_report.py
open output/dashboard.html

Project Structure

accessflow-analysis/
├── src/
│   ├── models/
│   │   └── impact_predictor.py    # XGBoost model training and evaluation
│   └── visualization/
│       ├── charts.py              # Static chart generation
│       └── dashboard.py           # HTML dashboard generator
├── notebooks/
│   └── accessflow_analysis.ipynb  # Interactive analysis notebook
├── fixtures/
│   └── *.xml                      # Toronto Open Data feed
├── docs/
│   └── adr/
│       └── ADR-003-*.md           # ML prediction target documentation
├── output/
│   ├── dashboard.html             # Interactive dashboard
│   └── charts/                    # Static PNG charts
├── generate_report.py             # Report generation script
└── requirements.txt               # Python dependencies

Key Results

Model	F1-Score	Accuracy
Baseline (always None)	0.2955	80%
Logistic Regression	0.8299	94%
XGBoost	0.9761	99%

Feature Importance

1. WorkPeriod (63%) — Schedule type is the strongest predictor
2. RoadClass (33%) — Road classification is the second strongest

Pedestrian Network Validation

High-impact closures have 44% more pedestrian infrastructure nearby:

Impact Level	Nearby Sidewalk
High	1,574m
Low	1,173m
None	1,064m

Dashboard

Open output/dashboard.html in a browser to view:

• Impact distribution charts
• Model performance comparison
• Feature importance analysis
• Confusion matrix
• Pedestrian network validation

Data Sources

Dataset	Source	Records
Road Restrictions	Toronto Open Data	1,834
Pedestrian Network	City of Toronto DAV	87,105 edges

License

MIT