Multi-Modal Mobility for Next Location Recommendation

This paper addresses the challenge of generalizing next location recommendations by leveraging multi-modal spatial-temporal knowledge. It proposes a novel method, M^3ob, that constructs a unified spatial-temporal relational graph (STRG) and employs a gating mechanism and cross-modal alignment to improve performance. The focus on generalization, especially in abnormal scenarios, is a key contribution.

• •Proposes M^3ob, a novel method for next location recommendation.
• •Leverages multi-modal spatial-temporal knowledge.
• •Employs a unified spatial-temporal relational graph (STRG).
• •Focuses on improving generalization, especially in abnormal scenarios.