Integration of Design of Experiments and Machine Learning

Researchers: Michelle Liu and Adrian Perez

This project explores how to combine the advantages of two different analytical approaches to improve decision making.  Design of Experiments (DOE) is used to create controlled experiments that minimize the data collection effort while maximizing the value of analysis results.  It originated in agriculture (to optimize factors that optimize farming operations) and is now a popular Six Sigma methodology in manufacturing and other industrial settings.  It stresses statistical efficiency but not robust visualizations.  Machine learning is typically associated with big data in business settings. It emphasizes the use of intuitive visualizations to support decision making.  The focus of the researchers’ effort will be on creating a standard controlled experimental design that would generate an intuitive classification tree.  The concept was tested by senior research associate Danielle Song, who created a classification tree for a group of supply chain capstone students working on the Watts Water Technologies project.