Profile Data in the Food Industry

• Activity Number: 307
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Section on Physical and Engineering Sciences
• Author(s): Eric Lagergren*+ and Wu Ge and Mahesh Padmanabhan
• Companies: Kraft Foods and Kraft Foods and Kraft Foods
• Keywords: Functional data ; Profile Data ; Chemometrics ; Principal Components ; Partial Least Squares ; Nonlinear Regression

Abstract

Profile data are increasingly common in the food industry. These data arise from a variety of sources including analytical chemistry measurements such as NIR, NMR, and LCMS, as well as temporal profiles such as 3-point bend texture and dough mixing curves. Kraft uses such data for calibration, prediction, and classification. NIR data are used to measure moisture in bakeries based on calibrations to vacuum oven measurements. Dough mixing curves are used to predict product characteristics, while NMR data are used to classify samples as same/different. We'll present some applications using profile data at Kraft and focus in greater detail on an application where the profiles are of systematically different lengths. Here typical Chemometrics tools do not strictly apply. Instead nonlinear parametric models are fit to the data with the estimated parameters used for subsequent analysis.