Amazon Last-Mile Routing Research Challenge

Implementation of the pool and select machine learning framework for solving the Amazon Last-Mile Routing Research Challenge.

Authors: Alexandre Florio, Paulo da Costa, Sami Serkan Özarık.

Description

This project contains all code developed by our group TurkishChurrasco for solving the Amazon Last-Mile Routing Research Challenge.

Context

Optimization of last-mile delivery routes is a complex operational task in transportation logistics. Many factors must be taken into account when designing efficient routes, including the total distance traveled by drivers, customer preferences and time windows, availability of parking spaces and spatiotemporal congestion patterns.

The pool and select framework is a machine learning-based framework for optimizing last-mile delivery routes. In the pool phase, a large number of candidate delivery sequences is generated based on structural information acquired from training data. In the select phase, the score of each candidate sequence is evaluated with a pre-trained and regularized regression model, and the sequence with the best (predicted) score is returned.

Reference

For more details on the framework and this implementation, please refer to:

Özarık, S.S., da Costa, P., & Florio, A.M. (2022). Machine Learning for Data-Driven Last-Mile Delivery Optimization. Available at SSRN: http://dx.doi.org/10.2139/ssrn.4012376

and

Florio, A.M., da Costa, P., & Özarık, S.S. (2021). A Machine Learning Framework for Last-Mile Delivery Optimization. In Winkenbach, M., Parks, S., & Noszek, J. (Eds.), Technical Proceedings of the 2021 Amazon Last Mile Routing Research Challenge (pp. XII.1–XII.19). MIT Libraries. https://hdl.handle.net/1721.1/131235

Dependencies

The implementation requires:

• The rapidjson (https://rapidjson.org) XML parser;
• A date.h implementation (e.g., this one);
• The mlpack library for training the score prediction models by Lasso regression.