CutScape: A Cycle-Based Optimization Framework for Partitioning Large-Scale Road Networks

Partitioning large-scale road networks is a fundamental problem that underpins scalable routing, transportation planning, network forecasting, and GeoAI applications requiring multi-scale representation and scalable learning over large graphs. Existing methods, however, often struggle to simultaneously achieve high partition quality, computational scalability, reproducibility, and flexibility across different partition objectives. This paper introduces CutScape, a cycle-based optimization framework for partitioning large-scale road networks, a representative class of sparse, planar-like spatial networks. Experiments on synthetic and benchmark road networks show that, compared with representative existing methods, CutScape achieves competitive or superior partition quality across objectives while scaling effectively to very large networks. The results demonstrate that CutScape provides an effective, flexible, and structurally interpretable framework for road network partitioning.

 

Terms of Use for Reproducibility and Non-Commercial Research Materials

Copyright (c) 2026. All rights reserved.

These anonymized data, source code, scripts, configuration files, and reproduction instructions are provided for scholarly peer review, verification, computational reproducibility, and non-commercial academic research in connection with the associated manuscript.

I. Purpose of Access

The materials may be accessed, inspected, compiled, executed, studied, and modified for the purposes of scholarly peer review, reproducing and verifying the findings reported in the associated manuscript, and conducting non-commercial academic research that builds upon, evaluates, compares, or extends the methodology.

The materials are not provided for general software use, product development, commercial application, commercial consulting, commercial services, or commercial redistribution.

II. No Open-Source License

The source code is not released under an open-source license. No open-source license, public-domain dedication, or implied license is granted.

Except for the limited permissions expressly stated in these terms, all rights are reserved by the copyright holder.

III. Permitted Uses

Editors, reviewers, and non-commercial academic researchers may use the materials to:

1. inspect the source code and documentation;
2. compile and execute the code;
3. reproduce the reported figures, tables, numerical results, and evaluation metrics;
4. verify the computational workflow described in the manuscript and accompanying reproduction instructions;
5. modify the code for non-commercial scholarly research, including evaluation, comparison, extension, and methodological development; and
6. publish scholarly findings derived from such non-commercial research, provided that the original materials are properly cited or acknowledged where appropriate and that modified source code or derivative software is not distributed for commercial purposes.

IV. Restrictions

Without prior written permission from the copyright holder, users may not:

1. use the materials, modified versions, derivative works, or further developments for commercial purposes;
2. incorporate the source code, algorithms, implementation, modified versions, derivative works, or further developments into commercial software, platforms, services, products, consulting work, or proprietary systems;
3. sell, rent, lease, sublicense, or otherwise commercially transfer the materials, modified versions, derivative works, or further developments;
4. redistribute the original source code, modified source code, derivative software, extensions, or further developments for commercial purposes;
5. remove, alter, or obscure copyright notices, license notices, patent notices, or attribution information.

V. Intellectual Property Notice

The methodology implemented in the source code may be the subject of pending intellectual property protection. No patent license or other intellectual property right is granted, except for the limited permissions expressly stated above for scholarly peer review, verification, reproducibility, and non-commercial academic research.

VI. No Warranty

The materials are provided "as is", without warranty of any kind, express or implied, including but not limited to warranties of merchantability, fitness for a particular purpose, accuracy, reproducibility, or non-infringement.

The copyright holder shall not be liable for any claim, damages, or other liability arising from use of the materials.

VII. Double-Blind Review

To preserve double-blind review, identifying information has been omitted from these materials. The copyright holder and any applicable institutional or contact information may be identified after completion of the review process, as appropriate.