MOVER: Mobility Modeling with User Variability and Enriched Representations

This is the Repo for our submission to ACM SIGSPATIAL Cup 2025: Human Mobility Prediction Challenge

Mover combines trip-aware segmentation, enriched location profiles, and cluster-based user adaptation.

1) Data Settings

Update dataset [ROOT] path in: ./datasets/builtin.py

• Download Trip Detection CSV files (replace the original compeition data)

Place Under:

[ROOT]

• Download Location Profile CSV files

Place Under:

./datasets/location_profile/v2/

• Download User Cluster CSV files

Place Under:

./datasets/user_cluster/

2) Download Model Weights

Place under:

./_runs/

3) Run Inference

bash ./_bash/inference.sh