Bayesian-Symbolic Integration for Uncertainty-Aware Parking Prediction
Summary: arXiv:2603.27119v1 Announce Type: cross
Abstract
Accurate parking availability prediction is critical for intelligent transportation systems, but real-world deployments often face data sparsity, noise, and unpredictable changes. Addressing these challenges requires models that are not only accurate but also uncertainty-aware. In this work, we propose a loosely coupled neuro-symbolic framework that integrates Bayesian Neural Networks (BNNs) with symbolic reasoning to enhance robustness in uncertain environments. BNNs quantify predictive uncertainty, while symbolic knowledge extracted via decision trees and encoded using probabilistic logic programming is leveraged in two hybrid strategies:
- Using symbolic reasoning as a fallback when BNN confidence is low.
- Refining output classes based on symbolic constraints before reapplying the BNN.
Evaluation and Results
We evaluate both strategies on real-world parking data under full, sparse, and noisy conditions. The results demonstrate that both hybrid methods outperform symbolic reasoning alone, indicating the efficacy of integrating BNNs with symbolic reasoning in dealing with uncertainty. Specifically, the context-refinement strategy consistently exceeds the performance of Long Short-Term Memory (LSTM) networks and BNN baselines across all prediction windows.
Significance of Findings
Our findings highlight the potential of modular neuro-symbolic integration in real-world, uncertainty-prone prediction tasks. The ability to effectively manage uncertainty in predictions is critical for applications in intelligent transportation systems, where decision-making relies heavily on accurate data regarding parking availability. This innovative approach not only enhances prediction accuracy but also offers a robust solution for handling the complexities associated with real-world data.
Conclusion
In summary, this work presents a novel framework for parking prediction that combines the strengths of Bayesian Neural Networks with symbolic reasoning. By addressing the uncertainties inherent in real-world scenarios, we provide a pathway for improving the performance of intelligent transportation systems. Future work will focus on further refining these methods and exploring additional applications in other domains where uncertainty plays a significant role.