Predicting Shelter Entry using Natural Language Processing of Homebase Case Notes

In collaboration with the Center for Innovation through Data Intelligence and the NYC Department of Homeless Services, this project uses Natural Language Processing of homelessness prevention case notes to predict an individual’s risk of shelter entry. Specifically, the study will investigate the ways unstructured case notes can be used to learn more about individuals using Homebase homelessness prevenetion services in New York City. Are there words, phrases, or topics that occur more frequently in the unstructured case notes of individuals who enter shelter after using Homebase services as compared to those who do not? And can a predictive model that assesses the probability of shelter entry based on structured data be improved by incorporating insights from unstructured case notes?

The Subway as Fourth Place: Anomie, Flânerie and the “Crush of Persons”

As part of research practicum course at Hunter in Fall 2016, we conceieved and conducted a mixed methods research project to assess social behavioir and interaction on the New York City subway. We collected more than 4,000 detailed observations of passenger behahvoir as well as in-depth “subway diaries” from eighteen research participants. Using logistic regression, we modeled the factors that influence how passengers direct their gaze and configure their bodies while riding the subway. The diaries helped us interpret and understand the observations.

The results of the study have been published in the peer-reviewed journal, Applied Mobilties. The research was also covered by the Daily News and CBS New York.