Spatio-Temporal Multivariate Weather Data Clustering Using DBSCAN And K-MEDOIDS Methods

Abstract

This thesis focuses on the examination of the efficacy of well-known data clustering techniques, namely DBSCAN and K-Medoids, in categorizing spatio-temporal multivariate weather data obtained from various disciplines such as atmosphericscience, Earth sciences, and environmental science. The data, which is generated through monitoring specific regions over a period of time, typically consists of four dimensions: time, longitude, latitude, and variables such as temperature and wind speed. The temporal dimension is used as the basis for clustering the data. The study proposes new quantitative metrics to evaluate the results of the clustering process. The findings indicate that while popular clustering algorithms are effective in handling simple synthetic data, they face challenges when applied to complex real-world data. Furthermore, the results show that as the number of variables in the dataset increases, the performance of the clustering methods worsens.