Keyword: Artificial Inteligence, Spatial Database, Spatial Modeling, Web Developement, Data Analysis
Project Markdown
2023
Research
The Indego Station Planner is a data-driven planning tool and machine learning model developed to support the expansion of Philadelphia’s Indego bike share network. The project combines predictive modeling with interactive mapping to help planners and residents evaluate where new stations should be located based on ridership potential, accessibility, and equity.
Although Indego provides more than two thousand bikes across over two hundred stations, much of the network remains concentrated in Center City.
Surrounding neighborhoods have limited coverage, leaving many residents without equal access to bike share infrastructure. The project was initiated by Philadelphia’s Office of Transportation, Infrastructure, and Sustainability (OTIS) and Bicycle Transit Systems (BTS) to guide Indego’s multi-year expansion plan and ensure it grows in a way that is both equitable and data-informed.
Eight machine learning models were trained and evaluated through spatial K-fold cross-validation, testing over 320 parameter combinations. Among these, the XGB-Poisson model achieved the highest accuracy, with a Mean Absolute Percentage Error of 46 percent and an average deviation of about 40 rides per grid cell per week. This model, integrated into the Station Planner web application, powers real-time predictions that allow users to test new station configurations, explore ridership outcomes, and save or reload their scenarios for future planning.
The Indego Station Planner was developed in Philadelphia, Pennsylvania, in 2023 as part of the Master of Urban Spatial Analytics (MUSA) practicum at the University of Pennsylvania. Working collaboratively with Rebekah Adams and Aidan Rhianne, under the guidance of Michael Fichman, Matt Harris, Elizabeth Delmelle, and Mjumbe Poe.
The tool continues to inform the city’s ongoing efforts to increase station density in high-ridership areas and extend service to underserved neighborhoods.
The XGB-Poisson model demonstrated robust accuracy in predicting ridership and successfully forecasted the performance of 23 new stations added in early 2023, with an average prediction error of only 19 rides per week. The Station Planner now provides city planners and community members with a shared, intuitive interface for designing an inclusive and efficient bike share network. By combining transparent data analysis with participatory design, the project advances the vision of a more connected, equitable, and sustainable Philadelphia.