Black-box modelling approaches for the prediction of microbiological bacterial growth

This paper presents two black-box modelling approaches for predicting bacterial growth curves; the two approaches are developed and compared by using Support Vector Machines (SVM): the first approach is aimed at predicting the parameters of an already existing model from the available measures, whereas, in the second approach, the resulting SVM itself plays the role of the model. The simulations are based on real experimental data, show that the two approaches have similar prediction capabilities but have different characteristics, which suggest the use of the most appropriate approach depending on the availability of a reliable parametric model.