nYDest

nYDest is a modular, open-source framework for modeling and forecasting urban destination choices, with a focus on the New York metropolitan area. It provides a data model, machine-learning components, and tooling to estimate the probability distribution over potential destinations given an origin, time of day, and contextual features such as weather, events, and transport availability. The goal is to support research, planning, and service design by producing interpretable destination forecasts.

The framework supports ingestion of diverse data sources, including anonymized trip records, public transit schedules (GTFS), points of interest, land use data, and geographic boundaries. The core modeling approach combines probabilistic methods with graph-based representations of the urban network, enabling the capture of routine behavior as well as network effects like travel time, connectivity, and accessibility. Outputs typically include destination probabilities, expected travel times, and actionable summaries for routing, scheduling, and capacity planning.

Architecture and interoperability are central features. nYDest is designed to be modular, allowing batch analyses or streaming predictions, and can run in Python environments on local servers or cloud platforms. It emphasizes data interoperability by supporting standard formats such as GeoJSON and GTFS, and it offers privacy-preserving options for aggregation and reporting to minimize exposure of individual travel patterns.

Development and usage: nYDest emerged from a collaborative urban analytics effort and has been adopted by researchers, city planners, and mobility providers for demand forecasting, urban planning, and transit optimization. While widely used in demonstrations and studies, it remains a community-driven project with ongoing updates and documentation.

See also: origin-destination modeling, mobility analytics, GTFS, GeoJSON.