Automated Classification of Bioprocess Based on Optimum Compromise Whitening and Clustering

J. Kukal, J. Mareš, J. Náhlík, P. Hrnčiřík and M. Klimt

Abstract
The proposed methodology of technological state classification is based on data smoothing, dimensionality reduction, compromise whitening, and optimum clustering. The novelty of our approach is in the stabile state hypothesis which improves initialization of c-mean algorithm and enables interleaved cross-validation strategy. We also employ the Akaike information criterion to obtain the optimum number of technological states that minimize it, but using as many as possible clusters and components. The general approach is applied to state classification of Pseudomonas putida fed-batch cultivation on octanoic acid.

This work is licensed under a Creative Commons Attribution 4.0 International License

Keywords
fed-batch cultivation, technological state classification, dimensionality reduction, clustering, stabile state hypothesis