Machine Leaning Methods
Multiple statistical methods can be used to set up a predictive model.
In simple situations with linear relationships and only a few possible
predictors, even linear models can be sufficient. Depending on the aim
of the model, fast and simple methods like PLS can be used to get a
rough idea of the prediction performance in process development. If the
model should be used in real-time in a production process, more advanced
non-linear methods such as Random Forest or Neural Networks should be
considered. Here we give an overview of the advantages and disadvantages
of some selected machine learning methods (Table 1).
Table 1: Most prominent
statistical methods used for prediction of quality and quantity in
bioprocess monitoring and control