2.5. Application of machine learning
There were four factors in these experiments, namely, drum speed (rpm),
drum surface temperature (°C), residual moisture content (% wb) and
lipid yield (mg g-1 of dry biomass). Residual moisture
content and lipid yield of corresponding drum speed and drum temperature
were mapped to get two features. The training data sets were labelled
good (1) or poor (2) based on the lipid yield and residual moisture
content (Table 1). Logistic regression with regularisation was used to
do the classification of the data gathered through experiments
(Dreiseitl & Ohno-Machado, 2002). The following steps mentioned in the
Figure 2 were employed.