Skip to content Skip to main navigation Report an accessibility issue

Shape/Sign Constrained Statistical Learning for Engineering Decision Making

Dr. Hoon Hwangbo
Postdoc Research Associate
Texas A&M University
Friday, March 2, 2018   2:30-3:30pm
JDT 410

Abstract: Data science or data analytics has demonstrated its strength in the evaluation and prediction of system behavior and has become essential in engineering decision-making. The success in system analysis, however, requires a deep understanding of engineering systems under consideration, in addition to the knowledge derived from the data. This is because a certain system property, be it mechanical, chemical, or biological, could restrict a system’s behavior but a purely data-driven approach overlooks such engineering constraints. In our research undertaking, we incorporate the pertinent yet imprecise engineering knowledge or physical constraints into a statistical learning process via imposing relevant shape/sign constraints on a system response function; the constraints include the requirements of positivity or negativity, monotonicity, convexity, or smoothness of the response function. Such shape/sign constrained learning, when applied to wind power production data, not only improves the estimation and prediction accuracy of the learning process, but enhances the interpretability of the resulting models and facilitates decisions for operations. The shape/sign constrained learning problem is in fact relevant to other engineering applications such as advanced manufacturing, which will be also discussed in the seminar.

Bio: Dr. Hoon Hwangbo is currently a postdoctoral research associate in the Department of Industrial and Systems Engineering at Texas A&M University. He received his Ph.D. degree in Industrial Engineering from Texas A&M University in 2017. His research interests are in the area of system informatics, and quality and data science, with a focus on shape-constrained statistical learning and nonparametric methods. His doctoral dissertation work was to develop data science methods for evaluating and improving the performance and operations of wind energy systems. He is a member of IISE and INFORMS.