Predicting cow milk quality traits from routinely available milk spectra using statistical machine learning methods

Numerous statistical machine learning methods suitable for application to highly correlated features, as those that exist spectral data, could potentially improve prediction performance over the commonly used partial least squares approach. Milk samples from 622 individual cows with known detailed protein composition and technological trait data accompanied by mid-infrared spectra were available assess predictive ability of different regression classification algorithms. The regression-based approaches (PLSR), ridge (RR), absolute shrinkage selection operator (LASSO), elastic net, principal component regression, projection pursuit spike slab random forests, boosting decision trees, neural networks (NN), a post-hoc approach model averaging (MA). Several (i.e., discriminant analysis (PLSDA), support vector machines (SVM)) also after stratifying traits interest into categories. In analyses, MA was best method 6 14 investigated [curd firmness at 60 min, αS1-casein (CN), αS2-CN, κ-CN, α-lactalbumin, β-lactoglobulin B], whereas NN RR algorithms 3 each (rennet coagulation time, curd-firming heat stability, curd 30 β-CN, A, respectively), PLSR pH, LASSO CN micelle size. When divided 2 classes, SVM had greatest accuracy majority investigated. Although well-established PLSR-based performed competitively, analyses reduced root mean square error compared between 0.18% (κ-CN) 3.67% (heat stability). use modern spectroscopy may some traits.