Textual Intelligence

3A Deep Learning Methodology with Transformers LLM to Calculate the Global Temperature Difference in Recent Years

Ana Carolina Borges Monteiro*, Reinaldo Padilha França^† and Rodrigo Bonacin^‡

Renato Archer Information Technology Center (CTI), Campinas, São Paulo, Brazil

Abstract

Climate change is a new challenge for humanity, so investing in climate change prediction technologies to alert governments and society regarding increases in annual global temperatures is an essential task. In recent decades, artificial intelligence has gained more prominence and more applicability in different sectors. Based on this, the present study aims to develop a neural network with Transformers LLM technology for predicting annual global temperature increases, based on the analysis of open-source climate data acquired from the year 1960 to the current period. As a result, the study presented a prediction of 1.17°Celsius in the global temperature for the year 2023. This value was compared with global temperature data for the year 2023 collected by renowned climate agencies, which corroborated the value predicted by the model developed in the present study. It is important to highlight that the model just as for the year 2023, there was a test loss of 0.05714299902319908 and consequently an MAE in the test of 0.2035137265920639, demonstrating the high reliability of the resulting data. The low loss rates and the high similarity with the temperature values predicted by climate agencies demonstrate ...