Modelling Directional Data by Bio-inspired Methods

Abstract

This thesis focuses on modeling wind direction data using von Mises and von Mises Mixture models. The Open-Meteo API was used to gather and process wind data for analysis. Traditional techniques were compared with bio-inspired optimization algorithms such as Particle Swarm Optimization, and other bio-inspired optimization algorithms were compared to each other. According to the results, PSO performed the best in terms of Negative Log-Likelihood, and mixture models are better at capturing intricate wind patterns. These results highlight that bio-inspired algorithms can improve the accuracy and robustness of modeling circular data compared to traditional approaches, make it become a useful tool in applications like live weather monitoring.