This is a project I worked on in summer 2019 under the supervision of Dr. Ayesha Ali (Professor of Statistics, University of Guelph). We extended the work of a previous PhD student who had used mid-infrared spectroscopy data on bovine milk samples to predict the composition of the milk (particularly fatty acid contents and milk fat globule size). Fatty acid content and fat globule size are factors which determine the overall quality of the milk, but traditional methods for measuring these quantities are slow and expensive. Spectroscopy data is relatively quick and cheap to obtain, so using spectroscopy to accurately predict the milk composition could save time and money.
The original model developed for predicting milk composition was done using partial least squares regression, but my advisor believed we could achieve better predictive power using more complicated regression methods. My job was to implement various regression methods using R and MATLAB, and compare and contrast them to partial least squares regression models which were originally created. A full report of this project is included below, and MATLAB code for this project can be found on my GitHub.
Cam Jakub 