Uncertainty Quantification via Spatial-Temporal Tweedie Model for Zero-inflated and Long-tail Travel Demand Prediction

TL;DR

The paper addresses zero-inflated and long-tail travel demand in high-resolution origin-destination matrices and the need for uncertainty quantification. It introduces the Spatial-Temporal Tweedie Graph Neural Network (STTD), which uses the Tweedie distribution with parameters $\mu$, $\phi$, and $\rho$ and a diffusion-based spatial-temporal encoder to model the density $f_{TD}(\cdot)$. Experiments on real-world CDP and SLD datasets across multiple spatial-temporal resolutions show that STTD achieves more accurate point estimates and calibrated predictive intervals than baselines such as STZINB and STNB, with notable improvements in MAE and MPIW. The work demonstrates the value of principled uncertainty quantification for transport planning, enabling more reliable dynamic resource management under sparse and skewed demand.

Abstract

Understanding Origin-Destination (O-D) travel demand is crucial for transportation management. However, traditional spatial-temporal deep learning models grapple with addressing the sparse and long-tail characteristics in high-resolution O-D matrices and quantifying prediction uncertainty. This dilemma arises from the numerous zeros and over-dispersed demand patterns within these matrices, which challenge the Gaussian assumption inherent to deterministic deep learning models. To address these challenges, we propose a novel approach: the Spatial-Temporal Tweedie Graph Neural Network (STTD). The STTD introduces the Tweedie distribution as a compelling alternative to the traditional 'zero-inflated' model and leverages spatial and temporal embeddings to parameterize travel demand distributions. Our evaluations using real-world datasets highlight STTD's superiority in providing accurate predictions and precise confidence intervals, particularly in high-resolution scenarios.

Figures (2)

• Figure 1: Framework of STTD model.
• Figure 2: Surface plots for learned STTD parameters ($\mu,\phi,\rho$) on CDPSAMP10 (left) and SLDSAMP10 (right) test sets.