Big Data Risk Analytics for Engineering Management and Public Policy

In this course, I will first cover the foundational principles in risk analysis and will then delve into methods used for developing quantitative risk analytics. Big data risk analytics leverages computational statistics and data mining to build predictive models based on large data sets to draw insights for engineering management and public policy application areas. As data becomes more prevalent across many different areas of importance in engineering, policy analysis, and management, predictive risk analytics is emerging as an increasingly important topic. This course assumes a working knowledge of probability and statistics and builds from this to introduce modern supervised and unsupervised learning techniques. Besides covering the foundations in risk analysis as well as the key principles for developing predictive risk models with high generalization performance, the course will also cover a range of semi- and non-parametric algorithms (e.g., generalized additive models, tree-based models, neural nets, support vector machines), meta-algorithms (e.g., boosting and bagging) as well as widely used unsupervised learning techniques such as PCA and clustering.

IE59000

Credit Hours:

3

Learning Objective:

After taking this class students should be able to:
  • Delineate the key elements in risk analysis
  • Us the R programming language for statistical analyses and inferencing
  • Assess models' generalizability, predictive accuracy and bias-variance trade-off
  • Develop regression or classification models to draw insights for engineering management and public policy applications using real data
  • Understand the underlying theory behind the most widely used algorithms in supervised and unsupervised learning that are covered in class

Description:

In this course, I will first cover the foundational principles in risk analysis and will then delve into methods used for developing quantitative risk analytics. Big data risk analytics leverages computational statistics and data mining to build predictive models based on large data sets to draw insights for engineering management and public policy application areas. As data becomes more prevalent across many different areas of importance in engineering, policy analysis, and management, predictive risk analytics is emerging as an increasingly important topic. This course assumes a working knowledge of probability and statistics and builds from this to introduce modern supervised and unsupervised learning techniques. Besides covering the foundations in risk analysis as well as the key principles for developing predictive risk models with high generalization performance, the course will also cover a range of semi- and non-parametric algorithms (e.g., generalized additive models, tree-based models, neural nets, support vector machines), meta-algorithms (e.g., boosting and bagging) as well as widely used unsupervised learning techniques such as PCA and clustering.
Spring 2021 Syllabus

Topics Covered:

Prerequisites:

Familiarity with computer programming (preferably R), and STAT 512, IE 23000&33000 or equivalent

Applied / Theory:

50 / 50

Textbooks:

Risk Analysis by T. Aven, Second Edition by Wiley
The Elements of Statistical Learning Theory by T. Hastie, R. Tibshirani, and J. Friedman, published by Springer. NOTE: You can download this book free at: http://www-stat.stanford.edu/~tibs/ElemStatLearn/
A Handbook of Statistical Analyses Using R, Second Edition, B.S. Everitt and T. Hothorn. CRC Press (2010).

Computer Requirements:

ProEd Minimum Requirements:

view

Instructor(s)

Roshanak Nateghi

Email
rnateghi@purdue.edu