Exploring Climate Interactions: A Fuzzy Rough Set Theory Approach

Abstract

Analysing climate change is challenging due to climate data’s intricate and dynamic nature. The primary issue is starting with high dimensionality. High dimensionality impacts the model’s performance, computation time, cost, and accuracy. Feature selection can be employed as a strategy to address the issue of dimensionality reduction, resulting in more precise insights and the identification of more explicit patterns. Various techniques are used for feature selection. Still, there is scope for progress in this field. This study uses fuzzy rough set theory (FRST) to perform feature selection in the analysis of climatic data. The dataset in the present study, obtained from Kaggle, is an authentic climate change dataset in the real world. FRST effectively addresses uncertainty and vagueness in climate data by identifying the most relevant temperature parameters and treating them as the deciding attribute. We identified 25 reducts from the original dataset using FRST. Compared to the original dataset, the best reducts had good classification accuracy. It indicates that FRST reducts preserve the essential features of the original climate data, assuring the reduced dataset’s integrity and relevance. FRST was more accurate than usual climate data analysis methods, proving its efficacy.