We propose FLLL3M--Federated Learning with Large Language Models for Mobility Modeling--a privacy-preserving framework for Next-Location Prediction (NxLP). By retaining user data locally and leveraging LLMs through an efficient outer product mechanism, FLLL3M ensures high accuracy with low resource demands. It achieves SOT results on Gowalla (Acc@1: 12.55, MRR: 0.1422), WeePlace (10.71, 0.1285), Brightkite (10.42, 0.1169), and FourSquare (8.71, 0.1023), while reducing parameters by up to 45.6% and memory usage by 52.7%.
Cite
@article{arxiv.2507.08843,
title = {Can We Predict Your Next Move Without Breaking Your Privacy?},
author = {Arpita Soni and Sahil Tripathi and Gautam Siddharth Kashyap and Manaswi Kulahara and Mohammad Anas Azeez and Zohaib Hasan Siddiqui and Nipun Joshi and Jiechao Gao},
journal= {arXiv preprint arXiv:2507.08843},
year = {2025}
}
Comments
Accepted in the 17th International Conference on Advances in Social Networks Analysis and Mining (ASONAM 2025), scheduled for 25 - 28 August 2025 in Ontario, Canada