Interpolating low-cost sensor data
In this project I used GPR (Gaussian process regression) to interpolate low-cost PurpleAir aerosol sensor data around the Southeast U.S. and to visualize it on a map of the area.
Athens PurpleAir calibration
In this project I gathered data from 13 low-cost PurpleAir aerosol sensors as well as from a regulatory monitor in Athens, GA. I then built a linear regression model to correct the low-cost sensor measurements to be more accurate.
Regression of NIR beer spectra
In this project I used a dataset consisting of NIR (near-infrared) spectra of 40 beer wort samples and the corresponding dissolved solids concentration measurements to build several regression models. The models were able to use the spectra to predict the dissolved solids to within 0.2 Platos (crazy unit used in brewing!).
Models include: LASSO (read my post about LASSO and Ridge regression), PLS (partial least squares) and GPR (Gaussian process regression).