Inteligência artificial para a previsão de capacidade produtiva: desenvolvimento de um modelo preditivo de machine learning para suporte à decisão em uma linha SMT

Abstract

This study applies Artificial Intelligence techniques, specifically Machine Learning, to forecast production capacity in an SMT manufacturing line of a company located in the Industrial Pole of Manaus (PIM). The organization’s traditional method relies on manual time measurements and static calculations, which limit precision and slow down the identification of bottlenecks and production planning. The aim of this research is to develop a predictive model based on the XGBoost algorithm and integrate it into a Power BI dashboard, providing a more reliable and data-driven tool to support decision-making. The study is conducted as a quantitative case study using real production and downtime records from May 2024 to September 2025. After data preprocessing, modeling, and evaluation using metrics such as MAE, RMSE, and R², the proposed model demonstrated higher accuracy compared to the traditional approach, better reflecting the operational behavior of the production line. The integration with the dashboard enabled interactive analyses by shift, month, and historical performance, enhancing visibility over capacity indicators.