Comparing the Performance of Multiple Linear Regression, Random Forest, and Artificial Neural Networks for the Prediction of Weather on Mars

Abstract

This thesis explores several machine learning methods for time series forecasting for weather prediction on Mars. The colonization of Mars has been proposed and funded by both public and private organizations like the National Aeronautics and Space Administration (NASA) and the aerospace corporation Boeing. The colonization of Mars has many challenges, one of which is the reliable prediction of weather. Traditional weather prediction techniques, such as numerical weather prediction, are not feasible on Mars given the lack of infrastructure needed for such powerful methods. In this thesis, several machine learning methods were implemented to circumvent these computational requirements: multiple linear regression (MLR), random forest (RFs), and artificial neural networks (ANNs). The work done for this thesis will inform the research questions of future atmospheric informaticians investigating the colonization of Mars and will serve as a strong baseline for model performance and methodology. Code and data are freely available at https://github.com/jfdev001/mars-ml-mtsu-honors-thesis.