Predictive Analytics for Public Policy
Concentration: Quantitative Methods
This class will discuss predictive analytics (also known as data mining) for public policy. The class will cover major families of predictive analytics models, but will also cover end-to-end business processes for building models, history of data mining, and policy implications (privacy, civil rights, etc.) of predictive analytics in the policy arena.
We will cover the mathematics of the models, although we will not delve into the math in detail. Instead, the focus is on understanding of what the methods do, what their practical advantages and disadvantages are, and where they might be most useful. Prior coursework in probability, statistics, and basic linear algebra (i.e., knowing what a vector and a matrix are) are prerequisites.
Class readings will be drawn from open-source materials on predictive analytics. The software we will use is KNIME, one of the leading open-source predictive analytics packages. Assignments will include several problem sets in which students will be put in the role of an analyst and asked to decide which models would best apply and why, in addition to applying those models directly. The final project will be a paper and a brief presentation in which students will present their results of applying predictive analytics to a public policy problem.