Comparison of Machine and Deep Learning Models for the Prediction of Land Degradation

Abstract

The primary purpose of this study was to develop and compare artificial intelligence algorithms to determine which gives the best predictions on variables related to land degradation. Data for this project was taken from satellite imagery and readings from ground stations. Data used included precipitation, temperature, and ground cover (EVI) readings. After comparing both the machine and deep learning methods it was found that overall machine learning vastly outperformed the deep learning models. In the end, random forest was the most accurate with a mean absolute percent error of 10.52%, and the top three models were all based on decision trees.