Foundations of Prescriptive Analytics

Instructors: Ugur Cetintemel - Serdar Kadioglu - {ugur, serdark}@cs.brown.edu
Teaching Assistants: Charlene Wang - hwang35@cs.brown.edu
Mailing Lists: {cs2951ostudent, cs2951oheadtas, cs2951otas}@lists.brown.edu
Class Hours: Tuesdays 4pm - 6:20pm
Class Room: CIT 506
Office Hours: Tuesdays 1pm - 2pm (Serdar CIT 317)
Syllabus: Course Syllabus
Piazza: Signup | Course Piazza
Academic Code: Academic Honor Code

Olympic Logo

We are undoubtedly in the middle of an Analytics Revolution that enabled turning huge amounts data into insights, and insights into predictions about the future. At its final frontiers, Prescriptive Analytics is aimed at identifying the best possible action to take given the constraints and the objective. To that end, this course provides students with a comprehensive overview of the theory and practice of how to apply Prescriptive Analytics through optimization technology. A wide variety of state-of-the-art techniques are studied including: Boolean Satisfiability, Constraint Programming, Linear Programming, Integer Programming, Local Search Meta-Heuristics, and Large-Scale Optimization.

The students are exposed to the industrially relevant software packages such as IBM Optimization Studio. The practical challenges encountered in implementing such systems are also explored. Additionally, the life-cycle of decision support systems is discussed and problems from real-life application domains such as planning, scheduling, resource allocation, supply-chain management, and logistics are addressed.

Course Objectives

The primary goal of this course is to introduce the fundamental ideas behind optimization technology to the extent that you can utilize this knowledge to build your own solvers based on various paradigms. Both complete and incomplete search methods, particularly tree-search and heuristic techniques will be covered in order to present different trade-offs. By the end of this course you will be able to transform a given optimization problem into analytical models with complementary strengths, and then, tackle it using off-the-shelf general purpose solvers and/or writing your custom solutions. This course shall also complement descriptive and predictive analytics as it connects data-centric approaches with their optimum decision-making counterpart.

Announcements