Automated resume screening using machine learning: an ensemble-based approach for efficient candidate selection

Large scale recruitment is time-consuming and results in inefficiencies where manual screening of resumes is used. This study proposed a novel Machine Learning (ML) based ensemble framework for resume classification to improve the process of recruitment. The research has a two-phase experimental design. First, several ML classifiers such as Logistic Regression, Decision Trees, Naive Bayes and Random Forest have been benchmarked to determine baseline performance. Secondly on the basis of these results, the best performing models were merged in a strategic manner to form ensemble framework using soft-voting mechanism. The proposed ensemble framework performed significantly better than standalone classifiers on a standard benchmark data set with an accuracy of 98.44% (5-fold CV mean: 97.47% ±1.00%). Paired T-tests ( p < 0.05) were used to ensure the statistical significance, which confirmed the reliability of the performance gains. In order to resolve the robustness gap in literature, the framework was tested on a new curated expert-labeled heterogeneous dataset ( N = 598) with a Zero-Shot testing protocol. The ensemble showed outstanding stability with a performance decrease of just 1.10%, and an F1-score of 0.99. Additionally, an interpretability audit was used to confirm algorithmic fairness, which guarantees the selection based on merit. Finally, a full functional web-based prototype was created to illustrate practical viability in real-time resume classification and recommendation. The results prove that the framework is not only highly accurate and statistically reliable, but also practically implementable for real-world talent acquisition.