This was a code I wrote for Kaggle's Africa Soil Prediction Challenge. The motivation of the challenge was to see if participants could generate a model to predict the nutrient content of African soils, which could then be used to improve sustainable farming plans.
The data available were infrared absorption spectra of soil samples as well as remote sensing data such as elevation, albedo etc.
The goal was to see how well can one predict five target values: SOC (Soil Organic Carbon), pH, Calcium, Phosphorous, and Sand content.