Um estudo sobre a aplicação do algoritmo Random Forest em problemas de classificação e regressão

Abstract

Machine Learning (ML) is a subfield of Artificial Intelligence (AI) that encompasses two main approaches: supervised learning and unsupervised learning. Within these, there are further sub- divisions, such as the division of supervised learning into classification and regression problems, each of which has its own specific algorithms for application and problem-solving. To facilitate ML, various algorithms can be cited, such as ANNs, RF, SVM, kNN, clustering, and many others. Each algorithm has unique characteristics, construction models, and applications, mak- ing them suitable for specific types of problems based on their learning framework. This study aims to conduct research on ML and some of its algorithms, with a particular focus on the RF algorithm, applying it to three datasets and analyzing the results. The development was carried out in R, using the Random Forest library (RF) available. Three datasets were pre-selected, all belonging to the same application domain but with distinct particularities, aiming to better tailor the function parameters of the library to each dataset. Finally, the results were evaluated based on predefined calculation metrics, considering the parameterization used for each execution. The analysis yielded effective and satisfactory results. [resumo fornecido pelo autor]