Abstract
<jats:title>Abstract</jats:title> <jats:p>X‐ray fluorescence (XRF) is a widely established analytical technique for elemental analysis whose applicability can be expanded to new horizons with the incorporation of machine learning (ML). Grounded on data‐driven pattern recognition, ML is exploited to study different objects through explicit or implicit correlations with XRF data. This chapter addresses the synergistic combination of XRF and ML from both theoretical and practical perspectives. After outlining the main fundamentals of XRF, we delve deeper into ML theory, focusing on supervised learning via Random Forests, Support Vector Machines, and Artificial Neural Networks, data treatment procedures, and robust training and validation strategies. The main strengths and limitations of the XRF + ML approach were pointed out and four case studies – nicotine detection in electronic cigarettes liquids, milk adulteration assessment, soil cation exchange capacity prediction, and banknote authentication – were developed as a practical way to showcase different study possibilities. Explainability tools were incorporated aimed at demystifying and clarifying the models' inner workings as well as different performance metrics were extracted to further evaluate their accuracy. In conclusion, XRF + ML brings a powerful set of analytical tools, whose reliable deployment depends on careful dataset design, rigorous validation, and robust interpretability checks to avoid spurious inferences.</jats:p>