CausalPOI: Spatio-Temporal Graph-Based Causal Modeling for Cold-Start POI Check-in Forecasting

2018-09-08 · Source: cs.AI updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Civic Technology & Smart Cities · Depth: Expert, extended

Summary

CausalPOI, a novel spatio-temporal graph-based causal representation learning framework, addresses the cold-start POI check-in forecasting problem, predicting future check-in patterns for newly introduced Points of Interest. Published in the Proceedings of the 32nd ACM SIGKDD Conference in 2026, CausalPOI models temporal evolution and functional interactions with nearby POIs in urban contexts. It utilizes a Spatio-Temporal Functional Interaction Graph (ST-FIG) to capture spatial proximity and semantic relationships, and a Causal Inference Module to simulate factual and counterfactual scenarios. Experiments on real-world SafeGraph datasets across four US regions (Northeast, Midwest, South, West) demonstrate CausalPOI's superior performance, achieving up to 57.8% RMSE and 34.3% MAE reduction compared to the best baseline, providing interpretable insights for urban intervention analysis.

Key takeaway

For urban planners or business operators assessing new Points of Interest, CausalPOI offers a robust method to forecast check-in volumes and estimate causal impacts. You should consider integrating causal inference with spatio-temporal graph modeling to move beyond correlation-based predictions. This framework provides more interpretable and actionable insights for site selection and resource allocation, especially when evaluating competitive or complementary effects of new establishments.

Key insights

CausalPOI forecasts cold-start POI check-ins by modeling causal effects through spatio-temporal functional interaction graphs.

Principles

• Causal modeling disentangles true effects from correlations.
• Functional interactions extend beyond geographic proximity.
• Counterfactual scenarios clarify intervention impacts.

Method

CausalPOI constructs paired treatment-control graphs, encodes POI textual tags with BERT, learns functional interaction strengths via contrastive pretraining, and uses GATv2 and GRU for spatio-temporal representation.

In practice

• Use BERT for POI textual tag embeddings.
• Pretrain category encoders with InfoNCE loss.
• Apply GATv2 for localized graph encoding.

Topics

• Cold-Start Forecasting
• POI Check-in Prediction
• Spatio-Temporal Graph Neural Networks
• Causal Inference
• Urban Computing
• SafeGraph Datasets

Code references

• ZZQ-NTU/CausalPOI

Best for: AI Scientist, Research Scientist, Machine Learning Engineer

Related on AIssential

• CausalPOI: Spatio-Temporal Graph-Based Causal Modeling for Cold-Start POI Check-in Forecasting — CausalPOI uses spatio-temporal causal modeling to predict check-ins for new POIs, outperforming correlation-based methods.
• Benchmarking Counterfactual Prediction in Epidemic Time Series with Time-Varying Interventions — A new benchmark provides realistic counterfactuals for epidemic time series, revealing challenges in causal inference methods.
• Interventional Time Series Priors for Causal Foundation Models — CausalTimePrior generates synthetic interventional time series data to train foundation models for time series causal inference.
• Correlation Is Not Enough: Embedding Human Metadata for Individual Causal Discovery — Pretrained biomedical language models struggle with cross-domain causal inference due to false embedding proximity, requiring specialized contrastive training.
• Causal Decomposition Analysis with Synergistic Interventions: A Triply-Robust Machine Learning Approach to Addressing Multiple Dimensions of Social Disparities — An extended causal decomposition analysis evaluates synergistic effects of multi-domain interventions using a triply robust machine learning estimator.
• Moment Matters: Mean and Variance Causal Graph Discovery from Heteroscedastic Observational Data — Separating mean and variance causal influences improves interpretability and intervention design in heteroscedastic systems.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.