Conference Paper
Neurosymbolic Approach for Travel Demand Prediction: Integrating Decision Tree Rules into Neural Networks
An IWCMC conference paper on improving travel demand prediction by injecting decision-tree-derived symbolic rules into neural network learning.
Abstract
Accurate travel demand prediction is essential for effective transportation planning, infrastructure development, and resource optimization. Traditional models often lack both interpretability and the ability to capture complex, nonlinear relationships across geospatial and socioeconomic variables. To address this, we propose a Neurosymbolic AI framework that integrates decision tree (DT) based symbolic reasoning with neural network (NN) learning for travel demand forecasting. The model leverages multisource data including geospatial, economic, and mobility datasets and embeds interpretable if-then rules extracted from DTs as additional features in the NN. Our experiments demonstrate that this hybrid approach improves predictive accuracy across key metrics, achieving up to 24.19% reduction in Mean Absolute Error (MAE), 1.79% improvement in R2, and 6.49% in Common Part of Commuters (CPC) compared to NN models. Rules extracted at finer variance thresholds (e.g., 0.0001) capture nuanced patterns and enhance model alignment with observed commuter behavior. By combining symbolic interpretability with neural generalization, the proposed method advances both the transparency and performance of travel demand modeling.
In brief
What This Work Does
The paper improves trip-demand forecasting by letting a neural network learn not only from raw data, but also from decision-tree rules that make important travel patterns easier to interpret.
Research impact
Why It Matters
The work demonstrates a practical way to bring rule-based interpretability into data-driven transportation demand models.
Paper at a glance
Research Scope
Neurosymbolic Demand Prediction Pipeline
Build Features
Uses multisource geospatial, economic, and mobility datasets to represent origin-destination travel demand.
Extract DT Rules
Learns interpretable decision-tree if-then rules that capture symbolic travel-demand relationships.
Select and Encode Rules
Uses variance thresholds to select useful rules and encode them as additional neural network inputs.
Evaluate Demand
Measures prediction quality with MAE, R-squared, and Common Part of Commuters.
Key Contributions
- Introduces a neurosymbolic travel demand prediction framework combining DT-based symbolic rules with NN learning.
- Encodes interpretable if-then rules as additional neural network features.
- Uses variance-based rule selection to balance model complexity, interpretability, and predictive value.
- Reports improvements over NN baselines across MAE, R-squared, and CPC.
- Demonstrates that symbolic interpretability and neural generalization can improve transportation forecasting together.
Method Components
Decision Trees
Provide human-readable if-then rules that summarize important demand relationships.
Neural Networks
Capture nonlinear patterns in transportation data and support flexible demand prediction.
Rule-Augmented Learning
Combines symbolic rule information with data-driven training to balance accuracy and interpretability.
Variance-Based Rule Selection
Controls rule granularity so the model can capture nuanced commuter patterns without overwhelming the neural learner.
Research Gaps
- Rule fidelity
- Model interpretability
- Spatial transferability
- Multimodal data
- Planner-facing explanations
Publication Details
- Type
- Conference Paper
- Venue
- IWCMC 2025
- Year
- 2025
- Pages
- 600-605
Research Topics
Links and Access
Citation
@inproceedings{acharya2025neurosymbolic,
author={Acharya, Kamal and Lad, Meet and Sun, Liang and Song, Houbing},
title={Neurosymbolic Approach for Travel Demand Prediction: Integrating Decision Tree Rules into Neural Networks},
booktitle={IWCMC 2025},
year={2025},
pages={600--605},
doi={10.1109/IWCMC65282.2025.11059465}
}