MobiML

Framework for machine learning from movement data

Development of this framework was inspired by https://github.com/wherobots/GeoTorchAI

For ML model documentation recommendations and corresponding template, see https://github.com/anitagraser/model-card-template

Installation

Note: As of today (2026-03-17), one of our main dependencies, pymeos, is not available on Windows. Therefore we recommend using MobiML on Linux.

Development installation

Install uv.

Clone this repository.

Set up the project:

uv sync

Run tests:

uv run pytest

In your application that uses mobiml, add these lines to the pyproject.toml file:

[tool.hatch.metadata]
allow-direct-references = true

and install

uv add  ../my/local/mobiml

For an introduction to uv, see e.g. the docs.

MobiML modules

MobiML contains various modules for learning and data preprocessing for movement data.

• datasets: This module contains classes for handling popular movement datasets.
• models: This module contains models for a variety of mobility-related ML tasks.
• preprocessing: This module contains tools to preprocess movement data to make it ready for ML development. Preprocessing tools always return a mobiml.Dataset object.
• samplers: This module contains tools for sampling movement data while accounting for its spatiotemporal characteristics.
• transforms: This module contains various transformation operations that can be applied to datasets. Transforms convert a mobiml.Dataset into a different data structure.

Documentation

Usage examples are provided in the examples directory, with instructions.

Included models

• GeoTrackNet -- Anomaly detection in maritime traffic patterns based on https://github.com/CIA-Oceanix/GeoTrackNet, as presented in Nguyen, D., Vadaine, R., Hajduch, G., Garello, R. (2022). GeoTrackNet - A Maritime Anomaly Detector Using Probabilistic Neural Network Representation of AIS Tracks and A Contrario Detection. In IEEE Transactions on Intelligent Transportation Systems, 23(6). arXiv:1912.00682
• Nautilus -- Vessel Route Forecasting (VRF) based on https://github.com/DataStories-UniPi/Nautilus, as presented in Tritsarolis, A., Pelekis, N., Bereta, K., Zissis, D., & Theodoridis, Y. (2024). On Vessel Location Forecasting and the Effect of Federated Learning. In Proceedings of the 25th Conference on Mobile Data Management (MDM). arXiv:2405.19870.
• SummarizedAISTrajectoryClassifier -- A basic example model implementing LogisticRegression for trajectory classification in a federated learning setting.

Publications

[0] Graser, A. & Dragaschnig, M. (2025). Learning From Trajectory Data With MobiML. Workshop on Big Mobility Data Analysis (BMDA2025) in conjunction with EDBT/ICDT 2025.

@inproceedings{graser2025learning,
  title={Learning From Trajectory Data With {MobiML}},
  author={Graser, Anita and Dragaschnig, Melitta},
  booktitle={Proceedings of the Workshop on Big Mobility Data Analysis (BMDA2025) in conjunction with EDBT/ICDT},
  year={2025},
  url={https://ceur-ws.org/Vol-3946/BMDA-6.pdf}
}

Acknowledgements

This work was supported in part by the Austrian FFG programme 'AI for Green 2023' under project No. FO999910218 (AI4PT) as well as by the Horizon Framework Programme of the European Union under grant agreement No. 101070279 (MobiSpaces).