Using Machine Learning and Small Area Estimation to Predicting Building-level Waste Generation and Recycling

Research Focus

Data-driven city operations

Primary Subject Area

Sustainability and waste

In New York City, it is estimated that the typical resident disposes of 15 pounds of waste at home and an additional nine pounds at the office or from commercial establishments each week (City of New York 2015). The more than six million tons of waste, including recyclables and organics, annually generated in NYC – and considerable amounts produced in cities throughout the world – create significant operational and disposal costs, negatively impact quality-of-life, and contribute to increasing carbon emissions. Waste management is a primary component of urban sustainability, and NYC has established an ambitious target of zero waste to landfills by 2030. Achieving this goal will require novel, data-driven strategies to measure and benchmark waste generation that can enable more efficient routing and collection systems, targeted outreach programs, and behavior change mechanisms. Similar approaches are being used for energy efficiency in buildings and urban water use (Kontokosta 2013, 2015; Kontokosta and Jain 2015).

Goals and objectives. Our goal is to model and predict building-level waste generation and recycling activity for each of the more than 900,000 residential properties in NYC.  Due to cost and feasibility constraints, the City of New York Department of Sanitation (DSNY) does not maintain data on individual building collection; rather, it tracks the waste collected over an entire truck route. Using robust regression modeling and machine learning prediction algorithms, we will pursue the following objectives:

  • Objective 1: Integrate and analyze data from multiple sources to estimate weekly waste generation rates at the building-level;
  • Objective 2: Validate model performance with DSNY through targeted, hand-sampling of building-level waste collection;
  • Objective 3: Establish citywide building-level generation estimates to create waste benchmarking and performance management metrics;
  • Objective 4: Apply model estimates to support operational efficiencies through optimized route planning and collection schedules and frequencies to maximize service delivery and truck tonnage collection targets.

Scalable impact. This research is expected have a substantial, scalable impact on solid waste management in cities through improved operational efficiencies and data-driven strategies to achieve longer-term landfill waste reduction and diversion goals.

Agency Partner

NYC Department of Sanitation

Team Members

Boyeong Hong, Xinshi Zheng, Geoff Perrin, Nick Johnson, Constantine Kontokosta

Funding

MacArthur Foundation

Research Team

Constantine E. Kontokosta

DIRECTOR, URBAN INTELLIGENCE LAB; Associate Professor of Urban Science and Planning; Director, Civic Analytics Program

Nicholas Johnson

Postdoctoral Associate

Boyeong Hong

PhD Candidate

Xinshi Zheng

Civic Analytics Graduate Student Fellow

Geoff Perrin

Civic Analytics Graduate Student Fellow

Constantine E. Kontokosta

DIRECTOR, URBAN INTELLIGENCE LAB; Associate Professor of Urban Science and Planning; Director, Civic Analytics Program

ckontokosta@nyu.edu

Prof. Kontokosta brings training urban planning, data science, economics, and systems engineering to the data-driven study of cities.

Constantine E. Kontokosta is an Associate Professor of Urban Science and Planning and Director of the Civic Analytics program at the NYU Marron Institute of Urban Management. He also directs the Urban Intelligence Lab and holds cross-appointments at the Center for Urban Science and Progress (CUSP) and the Department of Civil and Urban Engineering (CUE). He is affiliated faculty at the Wagner School of Public Service, Visiting Professor of Computer Science at the University of Warwick (UK), and a Senior Scholar at the New York Academy of Medicine. Previously, he served as the inaugural Deputy Director of CUSP and Assistant Professor of Urban Informatics at CUSP and CUE, where he was part of the Center’s founding leadership team and designed and launched the first graduate program in urban informatics. He is the founding Principal Investigator of the Quantified Community research initiative that integrates hyperlocal urban sensors with city-scale data analytics to understand neighborhood dynamics and well-being, and is one of the largest community-driven IoT projects in New York City. He is a 2017 recipient of the National Science Foundation CAREER Award for his research in urban informatics for sustainable cities.

Trained in urban planning and computational methods (Columbia), finance and economics (NYU), and systems science and engineering (UPenn), Constantine brings an inter-disciplinary perspective to urban science that integrates fundamental research with impact-driven, use-inspired needs. His work leverages large-scale data with computational methods to understand and drive change in energy efficiency and climate policy, neighborhood change and the impacts of urban development, and community-driven air quality monitoring and environmental justice. Recent projects include research with NYC311 and Kansas City to measure bias in citizen complaint reporting for predictive analytics; with a homeless shelter provider to apply machine learning algorithms to identify at-risk homeless families; and with the City of New York, Washington, DC, and the UN to leverage large-scale data analytics for building energy and climate policy. Constantine’s research groups – the Civic Analytics Program and the Urban Intelligence Lab – are motivated by a desire to bring evidence to policy-making, to democratize knowledge through information transparency, and to uncover discrimination and bias in data-driven decision-making.

Constantine’s research is funded by the National Science Foundation (NSF), MacArthur Foundation, Sloan Foundation, the U.S. Department of Transportation, the NYC Mayor’s Office of Sustainability, the Lincoln Institute of Land Policy, and the U.S. Department of Housing and Urban Development, among others, and he has received several honors for his work, including the IBM Faculty Award, the Google IoT Research Award, the UN Data for Climate Action Challenge Award,  the Goddard Junior Faculty Fellowship, the Charles Abrams Award for Social Justice Research, and NYU awards for Teaching Excellence and Outstanding Service. Constantine has published more than 70 peer-reviewed publications in leading academic journals and conferences – in fields ranging from urban planning to signal processing – and has two forthcoming books on urban analytics and data-driven climate action. His research has been featured in the New York Times, the Wall Street Journal, the Economist, FastCompany, CityLab, Wired, CNN, NPR, and other media outlets. He holds a PhD, M.Phil, and M.S. from Columbia University, a M.S. from New York University, and a B.S.E. from the University of Pennsylvania.

He serves or has served on committees and advisory boards at the National Academies, DARPA, the NSF Northeast Big Data Hub, the UNEP Sustainable Buildings and Climate Council, and the Royal Institution of Chartered Surveyors, and was the Vice Chair and a Commissioner of the Suffolk County (NY) Planning Commission. In addition to his academic work, Constantine is an accomplished entrepreneur.

Nicholas Johnson

Postdoctoral Associate

Nicholas.johnson@nyu.edu

Nicholas E. Johnson is a Postdoctoral Associate in Civic Analytics at the NYU Marron Institute of Urban Management.

He obtained his PhD in Computer Science/Urban Science at University of Warwick’s Institute for the Science of Cities in 2018. Previously, Nicholas received a Masters degree from NYU’s Interactive Telecommunications Program in 2013 centering his work on exploring the impact and pervasiveness of waste streams in urban environments through physical computing and interaction design.  He has launched several citizen science initiatives and continues citizen-driven research as an organizer for the Public Laboratory for Open Technology and Science.  Nicholas’s current research focuses on the design and development of cyber-physical systems for monitoring urban environments and data-driven analyses to understand urban phenomena including waste generation and urban mobility,

Boyeong Hong

PhD Candidate

boyeong.hong@nyu.edu

My research interests focus on how to apply urban informatics to real world problems in urban planning and operations.

I hold a master degree in Applied Urban Science and Informatics from NYU Center for Urban Science and Progress (CUSP). While at CUSP, I was a Graduated Research Assistant in Identifying E-Waste (Electronic waste) generation in New York City project in addition to working on data analytics for capital planning with NYC Department of City Planning as part of my capstone project. Most recently, I have been working at the Pratt Center for Community Development translating geospatial data into problem solving insight through GIS mapping and analysis. Prior to CUSP, I have participated in various research projects related to urban planning and data analytics in Seoul, South Korea. I have a Bachelor degree in Architecture from Yonsei University and a Master of City Planning degree from Seoul National University.

Xinshi Zheng

Civic Analytics Graduate Student Fellow

xz1845@nyu.edu

Xinshi is a Master student and a Civic Analytics Graduate Student Fellow in the Urban Intelligence Lab.

He holds a M.S. in Civil Engineering from University of Illinois at Urbana-Champaign (2015), and a B.Eng in Architectural Environment Engineering from University of Nottingham, UK (2013). His research interests include developing data science applications for urban infrastructure planning and optimization, as well as geospatial analysis.

Geoff Perrin

Civic Analytics Graduate Student Fellow

gtp232@nyu.edu

Geoff Perrin graduated from the University of Michigan with a BS in Economics and Mathematics, and is currently pursuing an MS in Applied Urban Science and Informatics at NYU’s Center for Urban Science and Progress. He’s built machine learning algorithms relating to various topics, from predicting Medicare fraud to predicting Levi’s jeans sales.