As urban commercial environments continue to evolve rapidly, accurately modeling the economic impact of point of interest (POI) entry is essential for supporting data-driven advertising allocation, merchant onboarding, and regional business decision-making. Existing return on investment (ROI) prediction methods mainly focus on individual ads or merchants, overlooking the interaction effects among POIs and failing to capture the causal influence of structural changes in urban commercial ecosystems. In this paper, we introduce a novel research problem -- regional ROI forecasting under POI entry, which aims to predict the future ROI trajectory of a commercial region after the introduction of a new POI by modeling temporal ROI evolution and competition-driven interactions among surrounding POIs. To address these challenges, we propose UrbanROI, an intervention-aware causal forecasting framework for region-level economic outcome prediction. UrbanROI leverages an Interaction-aware Regional Graph Module to explicitly model competitive relationships among POIs based on user request contexts, spatial proximity, and request-level commercial signals, and constructs structurally aligned treatment and control graphs to simulate factual and counterfactual regional states. Extensive experiments on large-scale real-world POI and advertising datasets demonstrate that UrbanROI consistently outperforms strong baselines in post-entry regional ROI forecasting. The results demonstrate its effectiveness in temporal economic modeling, inter-POI interaction representation, and intervention-aware forecasting, providing support for regional economic analysis and business decision-making in dynamic urban commercial systems.
Figures & tables
Figure 1. An overview structure of UrbanROI, which consists of two main components: (i) Interaction-aware Regional Graph Module , which constructs a region-level graph from historical user requests to explicitly capture competitive relationships among existing POIs; and (ii) Causal Inference Module , which builds structurally aligned treatment and control graphs to estimate the impact of a post-entry POI on regional ROI by modeling both factual and counterfactual scenarios.
City
Region
User Request
POI
Added POI
Shanghai
4,450
98,569
621,746
47,262
Beijing
3,902
47,263
481,512
58,043
Shenzhen
3,231
67,772
513,164
58,002
Hangzhou
2,492
51,182
335,530
36,116
Guangzhou
3,545
36,117
479,485
51,180
Table 1. Statistics of datasets
City
Metric
Informer
PatchTST
iTransformer
TimesFM
Moirai-MoE
GCIM
CauGramer
UrbanROI
w/o Graph
w/o Causal
Beijing
RMSE ↓
1.15 (± 0.05)
0.94 (± 0.03)
0.89 (± 0.04)
0.86 (± 0.05)
0.84 (± 0.03)
0.81 (± 0.04)
0.79 (± 0.03)
0.72 (± 0.02)
0.81 (± 0.04)
0.83 (± 0.03)
MAE ↓
0.69 (± 0.04)
0.56 (± 0.02)
0.53 (± 0.03)
0.51 (± 0.02)
0.50 (± 0.03)
0.48 (± 0.02)
0.47 (± 0.03)
0.42 (± 0.02)
0.48 (± 0.03)
0.50 (± 0.02)
MAPE (%) ↓
19.85 (± 0.31)
17.08 (± 0.27)
16.25 (± 0.32)
15.71 (± 0.24)
15.34 (± 0.29)
14.91 (± 0.22)
14.56 (± 0.26)
13.42 (± 0.19)
14.88 (± 0.25)
15.16 (± 0.29)
Shanghai
RMSE ↓
1.05 (± 0.04)
0.88 (± 0.05)
0.83 (± 0.03)
0.80 (± 0.04)
0.78 (± 0.03)
0.76 (± 0.04)
0.74 (± 0.03)
0.67 (± 0.02)
0.76 (± 0.03)
0.78 (± 0.04)
MAE ↓
0.67 (± 0.03)
0.55 (± 0.04)
0.51 (± 0.02)
0.49 (± 0.03)
0.48 (± 0.02)
0.46 (± 0.03)
0.45 (± 0.02)
0.40 (± 0.02)
0.46 (± 0.02)
0.48 (± 0.03)
MAPE (%) ↓
19.72 (± 0.35)
16.81 (± 0.26)
15.91 (± 0.30)
15.32 (± 0.25)
14.96 (± 0.28)
14.53 (± 0.21)
14.17 (± 0.24)
13.01 (± 0.18)
14.49 (± 0.27)
14.82 (± 0.24)
Table 2. Estimation and ablation study of regional ROI forecasting. We evaluate model performance using Root Mean Squared Error (RMSE), Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE). The results of UrbanROI are highlighted in bold. The last two columns of the table present the results of the ablation study. w/o Graph denotes the model that removes the Interaction-aware Regional Graph Module, and w/o Causal denotes the model that removes the Causal Inference Module. To ensure robustness, all metrics are reported as the mean and standard deviation over ten independent runs.
Figure 2. Parameter sensitivity of UrbanROI.
Method
CF RMSE ↓
Effect RMSE ↓
Sign Acc. ↑
Spearman ρ↑
Last-value
0.79
–
–
–
Temporal-only
0.69
0.071
71.2%
0.40
Zero-effect
–
0.096
–
–
UrbanROI
0.60
0.054
85.0%
0.66
Table 3. Averaged semi-synthetic effect recovery.
Real Intervention
Placebo Intervention
RMSE ↓
MAE ↓
MAPE (%) ↓
Avg. Uplift
RMSE ↓
MAE ↓
MAPE (%) ↓
Avg. Uplift
Beijing
0.72 ± 0.02
0.42 ± 0.02
13.42 ± 0.19
0.118 ± 0.011
0.86 ± 0.04
0.52 ± 0.03
15.76 ± 0.28
0.021 ± 0.008
Shanghai
0.67 ± 0.02
0.40 ± 0.02
13.01 ± 0.18
0.104 ± 0.010
0.81 ± 0.03
0.49 ± 0.02
15.33 ± 0.25
0.018 ± 0.008
Shenzhen
0.64 ± 0.02
0.39 ± 0.01
10.21 ± 0.16
0.097 ± 0.009
0.78 ± 0.03
0.47 ± 0.02
12.18 ± 0.21
0.016 ± 0.007
Hangzhou
0.66 ± 0.02
0.43 ± 0.02
12.48 ± 0.17
0.109 ± 0.012
0.80 ± 0.04
0.51 ± 0.03
14.77 ± 0.26
0.020 ± 0.009
Guangzhou
0.72 ± 0.02
0.45 ± 0.02
11.91 ± 0.16
0.113 ± 0.010
0.85 ± 0.04
0.54 ± 0.03
14.16 ± 0.24
0.019 ± 0.009
Table 4. Placebo test results of UrbanROI.
Appendix figures & tables4 assets
Supplementary material from the paper’s appendix.
Appendix
Type
Example
POI
id : 262...790,
online time : 20250107,
lat : 30240540,
lon : 120106426,
tags : {
name : 肯悦咖啡(杭州灵隐店),
Appendix
Table 5. Examples of POI and user request
Figure 3. Pooled pre-entry ROI trajectories of observed-entry and matched not-yet-treated regions across five cities. ROI is normalized by each region’s 30-day pre-entry mean. Lines and shaded areas denote group means and 95% confidence intervals, respectively.
Beijing
Shanghai
Shenzhen
Hangzhou
Guangzhou
Beijing
-
0.75 (± 0.04)
0.73 (± 0.05)
0.75 (± 0.03)
0.80 (± 0.06)
Shanghai
0.80 (± 0.05)
-
0.72 (± 0.04)
0.73 (± 0.03)
0.79 (± 0.05)
Shenzhen
0.81 (± 0.06)
0.74 (± 0.04)
-
0.75 (± 0.05)
0.78 (± 0.03)
Hangzhou
0.79 (± 0.04)
0.73 (± 0.03)
0.71 (± 0.04)
-
0.79 (± 0.05)
Guangzhou
0.80 (± 0.05)
0.74 (± 0.04)
0.72 (± 0.03)
0.73 (± 0.04)
-
Avg. Δ RMSE
+0.08
+0.07
+0.08
+0.08
+0.07
Appendix
Table 6. Pairwise cross-city generalization results of UrbanROI in terms of RMSE. Rows denote the source cities used for training and validation, while columns denote the target cities used exclusively for testing. The last row reports the difference between the average cross-city RMSE and the corresponding city-specific RMSE for each target city.
As urban environments continue to evolve rapidly, accurately modeling the dynamic behaviour of Points of Interest is essential for supporting data-driven urban planning and commercial decision-making. While recent advancements in spatio-temporal graph learning have improved POI forecasting, most methods rely on proximity-based graphs and correlation-driven modeling, which overlook the functional dependencies between POIs and fail to capture the causal effects of urban interventions. In this paper, we introduce a novel research problem -- cold-start POI check-in forecasting, which aims to predict the future check-in pattern of a newly introduced POI, by modeling its temporal evolution and functional interactions with nearby POIs in a structured urban spatial context. To address these challenges, we propose CausalPOI, a spatio-temporal graph-based causal representation learning framework. CausalPOI leverages Spatio-Temporal Functional Interaction Graph to model semantic and spatial relationships between POIs, and constructs structurally aligned treatment and control graphs to simulate factual and counterfactual scenarios. Extensive experiments on real-world SafeGraph datasets demonstrate that CausalPOI significantly outperforms state-of-the-art baselines across the board, validating its effectiveness in spatio-temporal forecasting, semantic interaction modeling, and causal effect estimation, providing a more interpretable and actionable foundation for urban intervention analysis. Source code is available at Github.
Zhaoqi Zhang, Miao Xie, Yi Li +3
Nanyang Technological University, Singapore, Singapore · China Agricultural University, Beijing, China · Meituan, Beijing, China
Predicting a user's next Point-of-Interest (POI) based on their historical check-in records is a fundamental task in location-based services. While recent methods incorporating large language models have shown strong reasoning capabilities and promising results, they typically formulate the prediction task as a one-step trajectory-to-location mapping problem, making predictions prone to shallow trajectory correlations and historical frequency bias. We argue that users rarely choose locations directly and instead, they usually first form a traveling intention and then accordingly select specific POIs. Motivated by this insight, we propose IntentPOI, a two-stage intention-guided reasoning framework. In the thinking stage, we infer users' intermediate intentions by incorporating historical mobility patterns, similar peer behaviors, and the temporal contexts. In the acting stage, we first construct a compact candidate pool, and then perform intention-guided reasoning to identify locations that best align with the inferred intention. By explicitly decoupling intention inference from location prediction, IntentPOI transforms the next POI prediction from direct trajectory matching into intention-guided reasoning. Extensive experiments on three real-world datasets demonstrate that IntentPOI consistently outperforms eleven state-of-the-art baselines.
Qingxiang Liu, Anqi Liang, Zhuoyang Jiang +5
The Hong Kong University of Science and Technology (Guangzhou) China · Shanghai Jiao Tong University China · The Hong Kong University of Science and Technology Hong Kong +2
Next Point-of-Interest (POI) recommendation ranks a user's likely next location based on check-in history. Most recent rankers compress the trajectory into a single user vector and score every candidate through the same representation, ignoring that every candidate carries geographic coordinates and that the relevance of a past visit depends on where the candidate is located. Target attention from click-through-rate prediction conditions the user representation on the scored item, but its operator was developed for web items without geometry. This paper proposes CaST-POI, a candidate-conditioned POI ranker that keeps a standard target-attention reader and adds two bucketised biases to the attention logits: one for the recency of each past visit and another for its distance to the candidate being scored. Different candidates therefore read the same trajectory with different attention weights, grounded in real geographic distance rather than inferred from item embeddings. Comparing against seven sequential recommenders trained with the same split on three datasets NYC, TKY and CA under per-user leave-one-out with full-vocabulary ranking, CaST-POI improves MRR over the strongest baseline by 5.1%, 7.6% and 14.5%. Holm-corrected paired tests confirm the gains on TKY and CA. Ablation study locates the gain in the explicit revisit gate and the candidate-relative spatial bias, whereas the candidate-independent temporal bias has no measurable effect on all datasets. Code is available at https://github.com/YuZhenyuLindy/CaST-POI.
Zhenyu Yu, Chunlei Meng, Yangchen Zeng +3
College of Computer Science and Artificial Intelligence Fudan University Shanghai, China · College of Intelligent Robotics and Advanced Manufacturing Fudan University Shanghai, China · School of Cyber Science and Engineering Southeast University Nanjing, China +2