Year: 2025
Toronto’s Line 1 (Yonge–University) subway is the busiest corridor in the TTC network, serving hundreds of thousands of riders daily. However, frequent delays — especially passenger-related incidents such as platform overcrowding, misuse of alarms, and unauthorized track entry — reduce reliability and increase operational costs.
This project develops a risk-aware, data-driven scheduling framework that combines:
- Historical Delay Data Analysis (City of Toronto Open Data)
- Time-Series Risk Forecasting using Facebook Prophet
- Mixed-Integer Linear Programming (MILP) optimization using PuLP in Python
Our strategy, called Modified Option 2, dynamically adjusts train frequency in response to forecasted delay risk. Results show:
- Average delay reduced to 3.14 minutes during peak hours
- Operational savings of approximately $6,600 per day
- Balanced trade-off between cost control and service quality
- Delay Attribution: Identify and classify major passenger-related delay types.
- Risk Forecasting: Predict hourly disruption risk using Prophet.
- Schedule Optimization: Use MILP to dynamically adjust train frequencies.
- Cost Control: Minimize operations cost while maintaining service standards.
- Scalability: Provide a framework adaptable for future TTC expansion.
TTC-Line1-Optimization/ │ ├── report/ │ ├── final_report.pdf # Formal written report │ └── code_dump.pdf # Extracted Colab/code PDF │ ├── notebooks/ │ └── Final_Analysis_TTC_line_1 (1).ipynb # Main analysis notebook │ ├── src/ │ ├── forecasting.py # Risk forecasting with Prophet │ ├── optimization.py # PuLP MILP optimization model │ └── utils.py # Helper functions │ ├── data/ │ └── README.md # Instructions for downloading TTC dataset │ ├── results/ # Forecast & optimization outputs │ ├── requirements.txt # Dependencies ├── LICENSE # MIT License └── README.md # This file
You do not need to install anything. Click the badge below to open the notebook in Google Colab:
Steps:
- Open the notebook using the badge above.
- Download the TTC delay dataset from Toronto Open Data.
- In Colab, upload the CSV into the
/data/folder (use the folder icon → Upload). - Run all cells: Runtime → Run all.
- View results inside the notebook (graphs, optimized schedules).
- Clone the repository.
- Install dependencies from
requirements.txt. - Place TTC delay dataset in
data/. - Run the Python scripts (
src/forecasting.py,src/optimization.py).
| Strategy | Avg Delay (min) | Daily Savings |
|---|---|---|
| Option 1 – Min Freq | 4.20 | $47.7k |
| Modified Option 2 | 3.14 | $6.6k |
| Option 3 – Max-Min | 3.00 | $0 |
Modified Option 2 provides the best balance of cost savings and service reliability.
- Deploy models in real-time with IoT sensors.
- Expand to other TTC subway lines.
- Test passenger behavior interventions (awareness campaigns, staffing).
If you use this repository, please cite:
Demoz, D.S.TTC Line 1 Delay Reduction and Scheduling Optimization Using Risk-Aware Forecasting and MILP. University of Ottawa.
This project is licensed under the MIT License – see the LICENSE file for details.