Seoul multimodal traffic flows forecasted for entire city districts
SeoulMMOD: A Large-Scale Multimodal Origin-Destination Flow Benchmark
Machine Learning
Summary
Predicting how many people move between city areas helps plan transportation better. Existing data usually covers only one or two travel types, which makes it hard to see full travel patterns. The authors provide a large dataset from Seoul showing hourly travel flows across six transportation modes over three years. They tested different forecasting methods and found that some models work well only at larger scales and that combining all modes for training doesn’t always help. This dataset and their tests create a reliable way to study and improve multi-mode travel forecasts citywide.
What this means in practice
- •For urban planners: Use multi-year, multi-mode travel forecasts across city districts to optimize transport infrastructure and service planning.
- •For traffic management teams: Improve traffic flow control by integrating forecasts of human movement for multiple transport modes in city sub-districts.
Authors
Taeyoung Yu, Seonbin Jo, Jiwon Kim, Junyoung Byun
Abstract
Origin-destination (OD) flow forecasting supports urban mobility decisions by predicting the volume and patterns of human movement across a city. Most existing public OD benchmarks provide only a partial view of citywide mobility, covering a single transport service or at most two modes. With partial mode coverage, observed demand changes are difficult to separate into shifts across transport modes and changes in total travel. We introduce SeoulMMOD, a large-scale public benchmark for citywide multimodal mobility in Seoul. SeoulMMOD provides three years of hourly OD flows estimating total mobility for 6 urban travel modes across 25 districts and 426 sub-districts, together with travel-time, travel-distance, calendar, rainfall, and point-of-interest (POI) signals. Using SeoulMMOD, we benchmark 14 forecasting baselines under a common protocol and analyse the effects of spatial scaling, joint versus mode-separate training, and cross-year generalization. Results show that several spatio-temporal models that work at district level do not scale to sub-district OD forecasting, and that naive joint multi-mode training does not consistently improve accuracy. Together, the dataset and benchmark establish a reproducible testbed for multi-mode, multi-year OD forecasting at the city scale.