Kamal Acharya

Ph.D. Candidate at UMBC, AI Researcher

Back to Publications

Conference Paper

Neurosymbolic Approach for Travel Demand Prediction: Integrating Decision Tree Rules into Neural Networks

2025 IWCMC 2025 DOI: 10.1109/IWCMC65282.2025.11059465

Travel Demand Prediction Neurosymbolic AI Decision Trees

DOI Cite

Abstract

This paper introduces a neurosymbolic framework for travel demand prediction that integrates decision-tree-derived symbolic rules with neural networks. It uses geospatial, economic, and mobility data to build predictive features, extracts interpretable if-then rules, and incorporates those rules into neural learning to improve accuracy while preserving interpretability.

Plain-Language Summary

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.

Why This Paper Matters

The work demonstrates a practical way to bring rule-based interpretability into data-driven transportation demand models.

Research Summary

This paper addresses travel demand prediction with a neurosymbolic approach. Standard neural networks can model complex relationships, but they are often difficult to interpret. Decision trees are more interpretable, but they may not capture all nonlinear structure in the data.

The proposed framework extracts symbolic if-then rules from decision trees and uses those rules as additional information for neural prediction. This creates a hybrid model that benefits from both interpretability and predictive flexibility.

The study is relevant for transportation planning because demand models influence infrastructure investment, resource allocation, and mobility policy. A more interpretable model can help planners understand why certain demand patterns are predicted.

Key Contributions

  • Combines decision tree rules with neural networks for travel demand prediction.
  • Uses symbolic rules as interpretable features in neural learning.
  • Evaluates prediction quality with MAE, R-squared, and Common Part of Commuters.
  • Shows how neurosymbolic modeling can improve both accuracy and interpretability.

Publication Details

Type
Conference Paper
Venue
IWCMC 2025
Year
2025

Authors

Acharya K., Lad M., Sun L., Song H.

Research Topics

Travel Demand Prediction Neurosymbolic AI Decision Trees

Citation

@inproceedings{acharya2025-neurosymbolic-approach-for-travel-demand-prediction-integrating-decision-tree-rules-into-neural-networks,
  title = {Neurosymbolic Approach for Travel Demand Prediction: Integrating Decision Tree Rules into Neural Networks},
  author = {Acharya K. and Lad M. and Sun L. and Song H.},
  booktitle = {IWCMC 2025},
  year = {2025},
  doi = {10.1109/IWCMC65282.2025.11059465}
}