Learning a Lot from Only a Little: Genetic Programming for Panel Segmentation on Sparse Sensory Evaluation Data

We describe a data mining framework that derives panelist information from sparse flavour survey data. One component of the framework executes genetic programming ensemble based symbolic regression. Its evolved models for each panelist provide a second component with all plausible and uncorrelated explanations of how a panelist rates flavours. The second component bootstraps the data using an ensemble selected from the evolved models, forms a probability density function for each panelist and clusters the panelists into segments that are easy to please, neutral, and hard to please.