ENBIS-19 Post-Conference Course: Functional Data Analysis

4 September 2019; 14:00 – 18:00

The half-day course is organised under the umbrella of the ENBIS-19 conference in Budapest. You can find more information about the conference here.

Functional Data Analysis

Machine Learning and DOE Analysis of Functional Data

Wednesday, 4th September 2019, 14:00-18:00, Room Dudich

Christopher Gotwalt, JMP Division of SAS Institute

Functional data has been present for quite some time in industrial data settings but there has not been off-the-shelf software to handle it properly.  JMP Pro 14 introduces the Functional Data Explorer platform, which provides an easy to use, point and click interface that makes modeling with functional data much more approachable.  There are many applications of functional data in industry.  Many of them fall into two broad categories: (1) machine learning using functional data from the signal streams generated from industrial processes as inputs to predict yield or manufacturing defects, and (2) the analysis of designed experiments where the response represents a curve or a series of repeated measurements.  In this workshop we will show how a technique called functional principal components is a useful statistical tool to solve these two problems.  In this course we will demonstrate how to use the Functional Data Explorer in JMP Pro to analyze both functional response designed experiments and fit machine learning models to batch process data.  These examples are from a variety of industries, from semiconductor and pharmaceutical manufacturing to the formulation of chemical products.

All participants will get free access to JMP Pro and are invited to bring their Windows or Mac computers to interactively explore the concepts of functional data analysis.

  1. Introduction To Functional Data Analysis
    1. Visualizing Functional Data
    2. Data Cleaning and Transformation
    3. Basis Functions
    4. Functional Principal Components
  2. Analysis of Functional Response Designed Experiments
    1. Optimizing a Milling Process
    2. Surface Tension DOE: Identifying Factor Settings to Match a Target Curve
  3. Functional Regressor Machine Learning
    1. Yield Prediction of a Fermentation Process
    2. Fault Detection in Semiconductor Manufacturing

Chris is the Director of Statistical Research and Development for JMP.  After graduating with a Ph.D. in Statistics from North Carolina State University in 2003, Chris joined JMP as a senior research statistician. Chris has made many contributions to JMP as a computational statistician including developing algorithms for optimal design of experiments, neural networks, time-to-event modeling, measurement systems analysis, linear mixed models, multivariate analysis, optimization algorithms, and text analytics.  Since 2007 he has led the analytical development of JMP and as Director has led successful development initiatives in manufacturing quality methodology, reliability modeling, consumer research, and data mining.  Chris is also an Adjunct Professor in the Statistics Departments of two universities, the University of Nebraska and North Carolina State University, where he oversees the dissertation research of Ph.D. students in the area of generalized linear mixed models, the analysis of large scale computer simulation experiments, and “Big Data” approaches to imputing missing values.