This year I ramped up the amount of machine learning I covered in my fall class at Penn. For the final project, I had my students recreate the phenomenally creative restaurant health inspection prediction project from Chicago. There were two components of the project – the development of a predictive model and then the design of an application to convert the predictive intelligence into a application that the health department could use to better allocate its limited inspection resources. The students don't estimate models anymore complicated than logistic regression, but they do spend a great deal of time constructing training and test sets and validating their models. Below is a video from two of the students in the class, Shruthi Arvind and Kristen Coe, presenting their health inspection app.

Recently, the Urban Institute called for the creation of "neighborhood-level early warning and response systems that can help city leaders and community advocates get ahead of (neighborhood) changes." Open data and open-source analytics allows community stakeholders to mine data for actionable intelligence like never before. The objective of this research is to take a first step in exploring the feasibility of forecasting neighborhood change using longitudinal census data in 29 Legacy Cities (Figure 2). The first section provides some motivation for the analysis. Section 3 provides results and the final section concludes with a discussion of community-oriented neighborhood change forecasting systems. Neighborhoods change because people and capital are mobile and when new neighborhood demand emerges, incumbent residents rightfully worry about displacement.