<p>Over recent decades, ultraviolet (UV) photodetectors have attracted substantial interest owing to their diverse applications across multiple fields. Photodetectors engineered from wide-bandgap semiconductors such as metal oxide (MO_X) offer effective UV light detection, as these materials can efficiently absorb and emit high-energy photons. Despite these advantages, achieving high responsivity and rapid response times remains a critical challenge in developing UV photodetectors. This research uses a comprehensive experimental dataset from prior studies to train machine learning (ML) algorithms to predict material selection and to assess MO_X UV photodetector performance. The developed random forest (RF) model demonstrates exceptional predictive accuracy, achieving <i>R</i>² values exceeding 99.99% for training, testing, and validation data and over 92.3% for entirely new datasets. Further, SHapley Additive exPlanations (SHAP) analysis is performed to evaluate the influence of key descriptors on device performance quantitatively. The application of an ML model can aid in the selection of appropriate materials by considering the effects of the properties of MO_X materials on UV device performance, making it easier to design materials than to synthesise a library of molecules experimentally. The current study establishes a robust framework where machine learning bridges empirical data and theoretical understanding, facilitating the rational design of advanced energy materials.</p> Graphical Abstract <p></p>