incomplete-learning

The paper proposes MARCUS, a missing-aware region representation model that treats missing data as contextual urban signals for rent prediction, achieving state-of-the-art performance with significant error reduction on real-world datasets.