Decision Analytics for Business and Policy

94-867 (Heinz) / 19-867 (Engineering and Public Policy) / 12-768 (Civil and Environmental Engineering)

(This course has recently been redesigned.)

This course is for advanced masters students in analytics, and assumes intermediate python experience and prior exposure to management science topics. It introduces modeling frameworks and computational tools to address complex decision-making problems that arise in policy and business. The course is organized by technical topics, and at the same time driven by many real world applications. We cover three modules: prescriptive analytics (linear, integer, nonlinear, and stochastic optimization), predictive analytics (regression, trees, deep learning, simulation, Markov chain), and dynamic decision making (learning and optimization, Markov decision process, simulation optimization, dynamic programming, and reinforcement learning). Roughly speaking, we take 4 weeks to cover each module, and each week introduces a different topic.

We motivate our technical discussions by a rich set of applications in lectures, recitations, readings, quizzes, assignments, and project. These applications are drawn from many areas: economics, machine learning, public policy, health care, public health, revenue management, supply chain, transportation, and internet. Given the fast pace of this course, we expect students to take an active learning role by participating in hands-on modeling and computational exercises throughout the semester. Coding will be done in Python for the most part. Students will learn to use a range of modern optimization and machine learning packages, and also develop their own algorithms in learning and optimization.

Prerequisites: (1) An introductory course in management science / operations research, (2) intermediate Python programming skills, (3) basic understanding of data and machine learning, and (4) mathematical maturity.

Decision Making Under Uncertainty

95-760 (Heinz)

This course provides an introduction to modeling and computational methods used by policy-makers, managers and analysts to support decision-making. The first half of the course focuses on deterministic optimization, and covers linear programming, network optimization and integer programming. The second half of this course introduces risk and uncertainty, and includes methods to characterize uncertainty and methods to optimize decisions under uncertainty. Examples are drawn from a variety of domains where these decision-making methods can provide value for business and policy, such as transportation, energy, health care, logistics, manufacturing, supply chain management, etc.

The readings, lectures, recitations, homework assignments and exams will help students develop modeling skills, computational skills and analytical skills. Modeling skills involve translating a problem into a well-defined mathematical framework, using little more than pen and paper. Computational skills involve solving models in Excel. Analytical skills involve critically interpreting a model and translating results into insights for decision-making.