M3TR: Temporal Retrieval Enhanced Multi-Modal Micro-video Popularity Prediction

Researchers developed M3TR, a new AI framework that uses temporal retrieval and multi-modal analysis to predict micro-video popularity with 19.3% better accuracy than existing methods. The system combines a Mamba-Hawkes Process module to model user feedback patterns with temporal-aware retrieval to identify historically relevant videos based on content and popularity trajectories.

• →M3TR achieves state-of-the-art performance, outperforming previous methods by up to 19.3% in normalized mean squared error for micro-video popularity prediction.
• →The framework addresses key limitations in existing methods by modeling user feedback as self-exciting events and incorporating temporal patterns into content retrieval.
• →The system combines multi-modal content analysis (visual, audio, text) with popularity trajectory matching to improve prediction accuracy.
• →The Mamba-Hawkes Process module captures long-range dependencies in user interactions like likes, comments, and shares.
• →Extensive testing on real-world datasets demonstrates significant improvements in long-term prediction challenges.