Comparative Evaluation of Neural Network Architectures for Generalizable Human Spatial Preference Prediction in Unseen Built Environments

2025-10-12

AI Summary

Existing human spatial preference prediction models exhibit limited generalizability to unseen built environments. Method: We propose a cross-layout transfer evaluation framework tailored to the built environment, constructing a synthetic pocket park dataset that integrates physical, environmental, and social features; we introduce a generalization scoring metric based on the Area Under the Precision-Recall Curve (AUPR) to systematically benchmark graph neural networks (GNNs), convolutional neural networks (CNNs), and feedforward neural networks (FFNs). Results: GNNs significantly outperform CNNs and FFNs on unseen layouts, demonstrating superior capacity for modeling structural relational patterns and achieving robust cross-scenario generalization. This work establishes a verifiable architectural selection criterion and evaluation paradigm for deployable, human-behavior-aware Cyber-Physical-Social Infrastructure Systems (CPSIS).

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

The capacity to predict human spatial preferences within built environments is instrumental for developing Cyber-Physical-Social Infrastructure Systems (CPSIS). A significant challenge in this domain is the generalizability of preference models, particularly their efficacy in predicting preferences within environmental configurations not encountered during training. While deep learning models have shown promise in learning complex spatial and contextual dependencies, it remains unclear which neural network architectures are most effective at generalizing to unseen layouts. To address this, we conduct a comparative study of Graph Neural Networks, Convolutional Neural Networks, and standard feedforward Neural Networks using synthetic data generated from a simplified and synthetic pocket park environment. Beginning with this illustrative case study, allows for controlled analysis of each model's ability to transfer learned preference patterns to unseen spatial scenarios. The models are evaluated based on their capacity to predict preferences influenced by heterogeneous physical, environmental, and social features. Generalizability score is calculated using the area under the precision-recall curve for the seen and unseen layouts. This generalizability score is appropriate for imbalanced data, providing insights into the suitability of each neural network architecture for preference-aware human behavior modeling in unseen built environments.

Problem

• Evaluating neural network architectures for spatial preference generalization
• Comparing model performance in unseen built environment layouts
• Identifying optimal architectures for transferable human behavior modeling

Innovation

• Graph Neural Networks for spatial preference prediction
• Convolutional Neural Networks for contextual dependencies
• Feedforward Neural Networks for generalizable layouts