Can We Predict Your Next Move Without Breaking Your Privacy?
This work addresses privacy concerns in mobility modeling for users and applications, representing an incremental improvement by combining federated learning with LLMs for a specific domain.
The paper tackles the problem of next-location prediction while preserving user privacy by proposing FLLL3M, a federated learning framework with large language models, achieving state-of-the-art accuracy results on multiple datasets, such as 12.55 Acc@1 on Gowalla, and reducing parameters by up to 45.6%.
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%.