In research and development, while numerous measurement variables can be obtained, the number of samples that can be evaluated is limited. To utilize such multivariate data in product development, it is crucial not only to assess predictive accuracy but also to visualize and interpret the differences between samples and use this information to inform future development decisions.
Our company is developing "FLAIRS," a sensing technology that utilizes multidimensional response information obtained from multiple light-emitting probes. We use JMP to perform data verification, principal component analysis, regression and discriminant analysis, and visualization of inter-variable relationships on the obtained sensor data, and we also utilize the model comparison function of JMP Pro as needed.
This presentation will focus on evaluating the compatibility of food products with seasonings and other condiments as a case study. By performing multivariate analysis on sensor responses obtained from each combination and visualizing and interpreting the differences in compatibility between combinations, we will demonstrate an analytical process that organizes the characteristics related to compatibility judgment and considers directions for product improvement.
We will also explain how to extend this analytical framework to quality evaluation, including differences in raw materials, formulations, products, processes, and lots. We will introduce the use of JMP to organize the characteristics of samples from multidimensional sensor data and connect this to hypothesis formation, planning of the next experiment, and decision-making in research and development.
