On New York University Week: Not all building-related issues are reported in cities, so how can machine learning help fill the gaps?

Daniel Neill, professor of computer science, public service, and urban analytics and Director of the Machine Learning for Good Laboratory, delves into this question.

Daniel B. Neill, Ph.D., is Professor of Computer Science, Public Service, and Urban Analytics at New York University, jointly appointed at NYU’s Courant Institute Department of Computer Science, Robert F. Wagner Graduate School of Public Service, and Center for Urban Science and Progress, where he directs the Machine Learning for Good Laboratory. Dr. Neill’s research focuses on developing novel machine learning methods for social good, with applications ranging from medicine and public health to urban analytics and fairness in criminal justice. Dr. Neill works extensively on developing new analytical methods to improve population health through predictive modeling, early event detection, causal inference, and targeting of interventions to reduce disparities. He received his M.Phil. from Cambridge University and his M.S. and Ph.D. in Computer Science from Carnegie Mellon University.

Machine Learning and Underreported Building Issues

Each winter, tens of thousands of New Yorkers report heating and hot water problems through New York City’s 311 system, prompting a process to restore these essential services. Yet, it is likely that many more New Yorkers experience major heating and hot water outages but do not report these serious quality-of-life issues through 311.

Using machine learning techniques, we identify New York City neighborhoods with high rates of unreported heating and hot water issues, and estimate which subgroups of the population are most likely underutilizing 311. To pinpoint both geographic areas and subpopulations, we use census data, 311 call logs, and information on each building’s structural characteristics.

We use these data to estimate which buildings are likely experiencing heating and hot water issues. Among buildings that do report to 311, we also estimate which ones are reporting at lower rates than expected. Together, these analyses help identify where building issues are going unreported, and where reported data may understate the severity of heating and hot water problems.

For example, we found that neighborhoods with a high concentration of elderly residents and limited English speakers are more likely to underreport heating and hot water problems. Additionally, many neighborhoods in Queens appear to be underreporting, while several areas in Upper Manhattan and the Bronx experience frequent heating and hot water issues, both reported and unreported.

These analyses provide useful information for NYC’s housing inspections process. We hope that they will assist city agencies and advocacy groups improving access to 311 and increasing services to communities whose quality of life is impacted by heating and hot water problems.

Read More:

Machine Learning For Good Laboratory

[Euclid] – Estimating reporting bias in 311 complaint data

The post Daniel Neill, New York University – Machine Learning and Underreported Building Issues appeared first on The Academic Minute.