Turnover em obras de montagem eletromecânica: aplicação de Machine Learning na gestão de pessoas

This study investigated the factors influencing employee turnover in electromechanical assembly projects, focusing on the application of statistical techniques and machine learning to support people management. The research used a dataset of 7,333 terminations that occurred between January 2023 and July 2025 in a Brazilian company from the sector, covering individual, organizational, and work-related variables. After preprocessing and exploratory data analysis, three classification models were applied: binary logistic regression (with stepwise selection), decision tree, and Random Forest (with grid search calibration). The results showed that factors such as percentage of leave days, tenure, project type, and state of residence had greater relevance for position replacement. The Random Forest model achieved the best predictive performance, with an AUC-ROC of 0.878 and a Gini coefficient of 0.756, while logistic regression stood out for the interpretability of its coefficients. The findings indicate that integrating statistical methods and machine learning contributes to anticipating turnover scenarios, supporting more effective retention and allocation strategies in human resources management for industrial projects.