Proposal for a machine learning-based implementation model for the recruitment of engineering professionals at a public university

Abstract

This research aimed to develop and evaluate a machine learning model to optimize the hiring process of engineering professionals at a public university by reducing evaluation time, errors, and subjectivity in CV screening. A quantitative, applied, quasi-experimental approach was used, combining TF-IDF natural language processing, KNN classification with a One-vs-Rest scheme, and experiments on three datasets (10, 20, and 30 CVs). Data from real recruitment processes were processed in Google Colab through cleaning, vectorization, training, and evaluation stages. The model achieved 82% accuracy in the test set, consistently prioritized candidates, and identified academic degree and professional experience as key factors. It reduced average analysis time from 15 to 2.5 minutes per CV and lowered the error rate to below 2%, while standardizing evaluation criteria. The results demonstrate that the model is efficient, objective, scalable, and suitable for institutional implementation.