TDToolkit Components
TDToolkit components work together to address challenges in travel demand research through data sharing, standardized evaluation metrics, and consistent protocols. Each component can be used independently or as part of an integrated workflow, ensuring reproducibility and comparability across different modeling approaches.
DCMBench
DataBaseline datasets and model benchmarks with classical discrete choice models. Provides standardized data preprocessing pipelines and reference implementations for traditional econometric approaches.
ML-DCM-Bench
BenchmarkingStandardized evaluation framework across machine learning and discrete choice models with reproducible train/test splits and consistent performance metrics. Ensures fair comparison between traditional and modern approaches.
TransitCast
BenchmarkingApplication of time series and deep learning models to transit ridership forecasting problems. Provides real-world evaluation scenarios for short-term public transit demand prediction under disruptions.
DNN-Econ
MethodologyEconomic metrics extraction including elasticities and welfare measures for both discrete choice models and deep neural networks. Bridges interpretability gap between traditional and ML approaches.
ASU-DNN
MethodologyHybrid DCM-DNN architecture combining utility-based choice modeling with deep learning flexibility. Provides interpretable yet powerful models that leverage strengths of both paradigms.