Evaluating Engagement in Job Offers: A Comparative Analysis of Predictive Models Based on Job Applications

This thesis was conceived as a milestone towards implementing a robust tool for evaluating engagement of real-world job offers and providing insights to enhance their attractiveness, aligning with the current hiring market needs identified by the collaborator and data provider, Inrecruiting. The goal is to explore classical Machine Learning techniques compared to statistical benchmarks, to improve the prediction of job application rates as a proxy for measuring job offers engagement. To test the hypothesis, it evaluates both the most straightforward approach—an unconditional benchmark average—and the simplest machine learning approach, linear models, as baselines for comparison with Random Forest, using Mean Absolute Error (MAE) and Mean Squared Error (MSE) metrics. The project analyzes, cleans, and preprocesses real job offer announcements, which include numeric, categorical, and text features, the latter with an NLP approach. Additionally, Monte Carlo Simulation (MCS) is employed to verify the models accuracy and the assess of the relative importance of each feature during training.