Applications of Hybrid and Explainable AI Models for Public Employment Services

Unemployment remains a pressing socioeconomic challenge, impacting both individuals and societies. While data-driven Public Employment Services (PES) have emerged to optimize employment support, the integration of Artificial Intelligence (AI)-based profiling remains limited. Existing PES implementations primarily rely on interpretable statistical models such as Logistic Regression, but these often lack the accuracy required for effective decision-making. More advanced models, like Gradient-Boosting Decision Trees, offer improved predictive performance but suffer from a lack of interpretability, requiring the use of eXplainable AI (XAI) techniques. This research aims to develop an XAI profiling and clustering tool for unemployed individuals within the Swiss labour market. The study applies both unsupervised models (DBSCAN and HDBSCAN) and supervised models (Decision Trees, Random Forests, XGBoost, and Cat-Boost). Model performance is validated using cross-validated performance metrics, while interpretability is ensured through SHAP analysis. Scientifically, this study bridges the gap between AI applications in PES and XAI. Unlike previous studies that focus solely on supervised machine learning techniques, this study integrates unsupervised learning methods and introduces the explainable hybrid ML models to PES.