Análise Comparativa de Pesquisas Eleitorais e Resultados Oficiais nas Eleições Brasileiras para Presidente e Governador com Uso de Machine Learning

Abstract

This article aims to compare the electoral polls and the official ballot results from the year 2022 using Machine Learning techniques, focusing on the presidential and gubernatorial elections in the state of Pernambuco. The study centered on the reliability of electoral polls and the discrepancies in relation to the official results, seeking to identify patterns and factors that influenced these differences. The research adopted a quantitative approach, using data from the Poder 360 portal and the Open Data Portal of the Superior Electoral Court (TSE). The methodology involved data preprocessing, implementation of the Random Forest model, and performance evaluation using metrics such as MAE (Mean Absolute Error) and RMSE (Root Mean Square Error). The results revealed significant discrepancies between poll data and official results, as well as the importance of certain features during model training.