MARCUS: Missing-Aware Region Representation with Contextual Urban Signals for Rent Prediction

Chenya Huang; Bin Liang; Zhidong Li; Yuxi Lu; Kunqi Li; Justin Wang; Fang Chen

cs.LGAug 19, 2026

MARCUS: Missing-Aware Region Representation with Contextual Urban Signals for Rent Prediction

Authors: Chenya Huang, Bin Liang, Zhidong Li, Yuxi Lu, Kunqi Li, Justin Wang, Fang Chen

Organizations: Data Science Institute University of Technology Sydney Sydney, Australia · The Property Investors Alliance Sydney, Australia

Abstract

Multimodal urban data has expanded the applications of urban region representation learning, such as functional zone identification and real estate appraisal, but also introduces challenges caused by data incompleteness. Existing studies usually handle missing data through imputation, treating missingness as noise while ignoring its potential semantic value. To address this issue, we propose MARCUS, a missing-aware region representation model that treats missingness as a contextual urban signal. MARCUS models missingness in three stages: Intra Learning jointly encodes observed features and missing patterns, Inter Learning estimates modality reliability to guide cross-modal interaction, and Fusion uses missing-aware and time-aware gating to generate the final region embedding. We apply MARCUS to rent prediction, a task with long-term trends and seasonal fluctuations, using real-world datasets from Sydney and New York. Experimental results show that MARCUS achieves state-of-the-art performance, reducing MAE by 51.35% on Sydney and 12.62% on New York compared with the best baselines. Additional experiments, including an imputation-based ablation study and randomized additional-missingness analysis, further demonstrate the effectiveness of the proposed method.

Figures & tables

Fig. 1: Additional missing ratio analysis on the Sydney dataset. The additional missing ratio denotes the proportion of originally observed data that are artificially masked. Naturally missing records in the original data are kept unchanged.

Datasets	Description	Sydney	New York
Regions	Partition scheme	SA2	ZIP code
	# Regions	373	177
Census data	# Surveys	19,750	85,154
	Missing ratio	0.54%	0%
POI data	# Records	145,548	20,570
	# Categories	13	13

TABLE I: Dataset Statistics

Fig. 2: Example of POIs in the Sydney Region

Fig. 3: Analysis and Visualization on Sydney Rent Data

Symbol	Description
$\mathcal{R}$	Set of regions (e.g., SA2)
$N$	Number of regions
$\mathcal{M}=\{P,C,F\}$	Modality set (POI, census, traffic)
$\mathbf{s}$	Missingness vector for modality $m$ at $(i,t)$ (flag, ratio, and missing-location encoding).
$\mathbf{h}$	Intra-modality embedding for modality $m$ at $(i,t)$
$\mathbf{H}$	Inter-updated embedding after cross-modal interaction

TABLE II: Frequently Used Symbols

Fig. 4: Overview of the Proposed MARCUS Framework

Model	Sydney			New York
Model	MAE $\downarrow$	RMSE $\downarrow$	MAPE(%) $\downarrow$	MAE $\downarrow$	RMSE $\downarrow$	MAPE(%) $\downarrow$
ZIB	126.46	158.65	24.05	558.94	862.07	16.14
LR	112.59	142.80	21.69	353.18	511.62	10.92
SVR	94.40	105.73	16.72	336.75	381.17	11.14
RF	84.46	96.61	15.19	187.42	261.18	6.04
FLEXIREG	77.18	88.83	14.18	160.10	239.55	5.14

TABLE III: Performance comparison on Sydney and New York

Model	MAE	RMSE	MAPE(%)
MARCUS-w/o-IAM	38.24	50.34	8.48
MARCUS-w/o-RW	40.24	50.63	7.96
MARCUS-w/o-FG	75.39	90.08	14.20
MARCUS-w/o-TG	68.86	84.27	12.87
MARCUS-w/o-M	70.15	85.09	13.24
MARCUS-Full	35.29	45.91	7.08

TABLE IV: Performance comparison of MARCUS variants.

Fig. 6: Sydney Rent Prediction System-Simple View

Fig. 7: Sydney Rent Prediction System-Professional View

Urban Representation Learning

See all 12 papers

UrbanVerse: Learning Urban Region Representation Across Cities and Tasks
Feb 17, 2026Fengze Sun, Egemen Tanin, Shanika Karunasekera +3Urban Representation Learning Geospatial Representation Learning
School of Computing and Information System, University of Melbourne, Melbourne, Australia · School of Computer Science and Engineering, University of New South Wales, Sydney, Australia
When Correlations Mislead: Confounder-Aware Multi-View Urban Region Representation Learning
Sep 14, 2026Sean Bin Yang, Ying Sun, Zongyi Xu +6Causal Representation Learning Graph Representation Learning
Aalborg University, Aalborg, Denmark · Chongqing University of Posts and Telecommunications, Chongqing, China · East China Normal University, Shanghai, China
Rethinking Multi-View Fusion for Urban Representation Learning
Oct 22, 2026Sean Bin Yang, Zongyi Xu, Jilin Hu +5Urban Representation Learning Multi-View Feature Fusion
Aalborg University, Denmark · Chongqing University of Posts and Telecommunications, China · East China Normal University, China

Feb 17, 2026J/K move · Enter open · S save

UrbanVerse: Learning Urban Region Representation Across Cities and Tasks

Fengze Sun, Egemen Tanin, Shanika Karunasekera, Zuqing Li+2

School of Computing and Information System, University of Melbourne, Melbourne, Australia · School of Computer Science and Engineering, University of New South Wales, Sydney, Australia

Urban Representation Learning Geospatial Representation Learning Spatial Generalization Multi-Task Learning Graph Representation Learning

Open paper AskPDF arXiv

Recent advances in urban region representation learning have enabled a wide range of applications in urban analytics, yet existing methods remain limited in their capabilities to generalize across cities and analytic tasks. We aim to generalize urban representation learning beyond city- and task-specific settings, towards a foundation-style model for urban analytics. To this end, we propose UrbanVerse, a model for cross-city urban representation learning and cross-task urban analytics. For cross-city generalization, UrbanVerse focuses on features local to the target regions and structural features of the nearby regions rather than the entire city. We model regions as nodes on a graph, which enables a random walk-based procedure to form "sequences of regions" that reflect both local and neighborhood structural features for urban region representation learning. For cross-task generalization, we propose a cross-task learning module named HCondDiffCT. This module integrates region-conditioned prior knowledge and task-conditioned semantics into the diffusion process to jointly model multiple downstream urban prediction tasks. HCondDiffCT is generic. It can also be integrated with existing urban representation learning models to enhance their downstream task effectiveness. Experiments on real-world datasets show that UrbanVerse consistently outperforms state-of-the-art methods across six tasks under cross-city settings, achieving up to 35.89% improvements in prediction accuracy.

Abstract

Urban Representation Learning

Explore similar work

UrbanVerse: Learning Urban Region Representation Across Cities and Tasks

When Correlations Mislead: Confounder-Aware Multi-View Urban Region Representation Learning

Rethinking Multi-View Fusion for Urban Representation Learning