Data analysis - from scan to prediction - SoilCares

Machine learning is, in essence, the process of applying algorithms to identify patterns in the data that correspond with the ground truth of that data. In our case, the ground truth is the reference values found in the GSL, and the patterns are the spectra for each sample. The regression model calculates a function to transform a spectrum into each of its reference values. For example, the presence of a significant peak in the spectrum could correspond with a high Potassium concentration. We also create regression models that allow us to predict how confident the predictions for a spectrum are.