This dissertation addresses the innovative application of machine learn ing in predicting failures in power transmission assets. Emphasizing improve ments in accuracy and reliability compared to traditional methods, the work introduces machine learning techniques using the Random Forest algorithm in a sector that has historically been conservative in adopting this type of computational technology. The document is structured to include a theoretical foundation, relevant previous works, presented results, and concludes with directions for future research. Additionally, it discusses an approach of best choice of machine learning algorithms by the minimum sample size of ex amples, offering a tool developed to support decision-making. Through this academic endeavor, the dissertation aims to contribute to the technological advancement of the electrical sector.