Dr. Ken Steif is the Program Director of the Master of Urban Spatial Analytics program (MUSA) at Penn and has been teaching for more than a decade. He teaches exactly what he practices – at the intersection of public policy and data science. Currently he is teaching three graduate courses in the City Planning Department at the University of Pennsylvania.
MUSA/Smart Cities Practicum
This is a capstone course for MUSA and data-driven City Planning students interested in applying their data science skills to develop algorithmic solutions for governments around the country. Students learn how to wrangle and integrate often messy, across agency administrative data; how to mine those data for useful predictive features; how to interpret the results of contemporary machine learning algorithms and most important, how to communicate algorithmic solutions to non-technical decision makers.
Working in domains as diverse as housing, transportation, public health and public safety, students develop a markdown complete with source code to enable other cities, collecting the same data, to replicate their analysis. They also develop a web-based dashboard or mapping tool to help policy makers interact with their predictions. The projects are exhibited on this website.
‘Public Policy Analytics’:
This course teaches advanced GIS functionality and spatial analysis in the realm of urban planning. The class focuses on real-world GIS applications and, in combination with introductory statistics, provides students a framework for understanding how to efficiently allocate limited resources across space. Specific applications include network analysis; retail site-suitability; point-processes; spatial housing market analysis; predictive modeling; remote sensing, land use planning and more. The format of the class includes weekly lectures/in-class demos and weekly homework assignments. Check out the list of topics from this year’s course:
Planning Transit-Oriented Development
Comprehensive planning: The case of Urban Growth Boundaries
Modeling urban growth/environmental protection
Explanatory modeling: What’s the value of transit proximity?
Churn prediction
Predictive modeling competition: Hedonic home price regression
Network Analysis & Planning for disaster response
Remote Sensing: Detecting vacant land in Detroit
Business siting: Predicting success over space
Huff Modeling: Retail site suitability
Geospatial risk prediction (predictive policing)
Algorithmic bias
Machine learning and restaurant health inspection prediction
‘Land Use & Environmental Modeling’:
Urban and environmental planners are using spatial data and increasingly sophisticated empirical models to analyze existing patterns, parameterize key trends and processes, forecast alternative futures, and visualize key results for non-technical decision-makers.
This course focuses explicitly on these themes. By the end of the course, students will understand how to parameterize spatial data specifically for modeling as well as how to take top down and bottom up approaches to modeling comparable phenomenon. This is a GIS-oriented class that will rely on statistical modeling. We will use a variety of software packages including ArcGIS, R, the HEC Suite of hydrological tools and NetLogo.
McHargian overlay and data visualization
Fragmentation stats in R
CNgrid: How land use & soil type affect drainage
Watershed delineation in ArcHydro & HEC-HMS
Modeling floods in HEC-RAS
Modeling FEMA Hazus flood damage functions in R
Machine learning and flood inundation prediction – remote sensing
Agent-based modeling – modeling gentrification
Machine learning and urban growth models