From Data to Delivery: Data-Driven Engineering of Last-Mile Optimization for Quick Commerce in Rome and Montreal

This repository contains the full code, data, and model artifacts for my master's thesis project at LUISS Guido Carli. The project focuses on engineering and benchmarking advanced courier assignment strategies for last-mile delivery in quick commerce, using data-driven and machine learning approaches.

---

Project Structure

Code/ # All simulation engines, assignment strategies, model training scripts, and utils Data/ # Datasets and trained model files for reproducibility /Montreal # Cleaned Montreal datasets /Rome # Cleaned Rome datasets /ANFIS # Trained ANFIS models, scalers, and transformers Maps/ # GIS and visualization outputs

---

Getting Started

1. Clone this repository

   git clone https://github.com/DavidPaquette99/From-Data-to-Delivery-Data-Driven-Engineering-of-Last-Mile-Optimization-for-Quick-Commerce-in-Rome-.git
   cd From-Data-to-Delivery-Data-Driven-Engineering-of-Last-Mile-Optimization-for-Quick-Commerce-in-Rome-

2. Install dependencies

   pip install -r requirements.txt

3. Run a sample simulation

   cd Code/simulation
   python Run_Simulation.py

   (You may need to adapt paths depending on your use case. See Code/README.md for details on modules.)

---

Data

• All cleaned and processed data for Montreal and Rome are provided in Data/Montreal/ and Data/Rome/.
• Trained ANFIS models and scalers are in Data/ANFIS/ and used for courier assignment and regression models.

No sensitive or private data is included. For full raw datasets or Google API keys, please contact the author.

---

Folder Guide

• Code/ - Main simulation code, assignment strategies, training pipelines, utilities
• Data/Montreal/ - Cleaned and engineered data for Montreal
• Data/Rome/ - Cleaned and engineered data for Rome
• Data/ANFIS/ - Trained ANFIS models, scalers, and transformers
• Maps/ - Map visualizations

---

Reproducibility

• All code and data required to reproduce core results are included.
• To retrain models, see scripts in Code/model_training/.
• Pretrained models are loaded from Data/ANFIS/ (default paths are absolute, but can be adapted).

---

Dependencies

Major dependencies (see requirements.txt):

• Python 3.8+
• numpy
• pandas
• scikit-learn
• joblib
• osmnx
• matplotlib

---

Thesis Report

The full thesis report is available here:
David Paquette Master’s Thesis (PDF)

---

Citation

If you use this code or data, please cite:

Paquette, D. (2025). From Data to Delivery: Data-Driven Engineering of Last-Mile Optimization for Quick Commerce in Rome and Montreal. Master’s Thesis, LUISS Guido Carli.

---

Contact

Questions, feedback, or collaboration proposals welcome!
David Paquette
d.paquette@studenti.luiss.it