Optimizing Genetic Algorithm Parameters for Atmospheric Carbon Monoxide Models

Optimizing Genetic Algorithm Parameters for Atmospheric Carbon Monoxide Models

PROJECT NUMBER: 1 | AUTHOR: Meera Duggal​, Applied Mathematics and Statistics

MENTOR: Dorit Hammerling, Applied Mathematics and Statistics

ABSTRACT

The primary source of atmospheric carbon monoxide (CO) variability in the Southern Hemisphere is large burn events, making CO a useful proxy for fires. Therefore, predictive CO models over fire regions can help countries prepare for unusually large fire seasons. Fires are related to the climate through fuel dryness and availability, which varies with climate variability. Climate indices are metrics that summarize climate variability through changes in sea surface temperature and wind. We created a multiple linear regression model that uses these climate indices to predict atmospheric CO, as well as the R package \\textit{regClimateChem} to perform variable selection. This package offers three different variable selection techniques: stepwise selection, a genetic algorithm, and an exhaustive search. The exhaustive search always finds the best possible model but is computationally expensive. Stepwise selection runs quickly and is scalable but often fails to find the best model. We implemented a genetic algorithm as a compromise between computational expense and model accuracy. As a stochastic variable selection technique, the genetic algorithm has many parameters that affect the models it selects. Here we present a parameter optimization study for the genetic algorithm, seeking to balance computational expense and model quality. We find that a certain combination of parameters, for a four covariate case, results in 11.8% runtime saving and only compromises 0.3% accuracy compared to the default settings. In the five covariate cases, it was found that the optimized genetic algorithm trends towards the exhaustive method, on a high-performance computer.

PRESENTATION

AUTHOR BIOGRAPHY

Meera Duggal is a senior studying Applied Mathematics and Statistics. She is doing research in the Statistics department with Dr. Hammerling and William Daniels. The study that she performed was a genetic algorithm optimization for atmospheric carbon monoxide models. During her first year, she focused on optimizing the genetic algorithm for a four covariate model. This last semester she then learned how to use an HPC system to optimize the genetic algorithm for a five covariate case. In the following year, she will be doing her master’s in Statistics.