Estudio comparativo y aportes de los algoritmos de Machine Learning aplicados a los procesos de contratación en QA

Abstract

This study comprehensively compares different machine learning algorithms to assess their applicability in Quality Assurance (QA) contracting procedures. The evaluated algorithms encompass Nearest Neighbors, Linear SVM, Radial SVM, Gaussian Process, Decision Tree, Neural Network, Logistic Regression, Naive Bayes, and QDA. Following the Knowledge Discovery Database (KDD) process, the methodology includes a diverse set of evaluation metrics such as F1-Score, recall, accuracy, and AUC-ROC, as well as learning curves, boundary maps, confusion matrices, Matthews Correlation Coefficient (MCC), and complexity curves. According to the Gartner Magic Quadrant assessment, the results suggest that Neural Network, QDA, and Gaussian Process models exhibit strong performance and thorough evaluation, making them optimal for the case study presented in the paper. In contrast, Nearest Neighbors and Linear SVM models are considered suboptimal, indicating an opportunity to explore the reasons behind their behavior in the case study and how to modify them for improved results. The other algorithms also present various possibilities for adaptation to the current case study, either as models with limited analysis or as imprecise models that can offer valuable insights for future work on optimizing them more effectively. This study significantly contributes to advancing machine learning applications in recruitment procedures.