<p>This study examined how machine learning can be used to predict the formation of reverted austenite in 18% Ni maraging steels during overaging heat treatments. This transformation that plays a crucial role in determining strength and toughness. To achieve this, a dataset was developed by combining published results with new experimental measurements obtained through x-ray diffraction (XRD), enabling a detailed analysis of how temperature, time, and alloy composition affect the reversion process. Different regression models were tested, with the ensemble-based Voting Regressor showing the best performance, achieving strong agreement with experimental data (R² ≈ 0.98). Some discrepancies arose for very high temperatures and long heat-treatment durations, indicating the need for additional data for these conditions. The results showed that temperature is the primary factor controlling austenite reversion, while aging time has a smaller but noticeable influence. Alloying elements also played an important role, particularly titanium, which slowed reversion by forming Ni₃Ti precipitates that reduce the amount of nickel available in the steel matrix. These findings were reinforced by correlation maps and statistical analyses, which aligned well with long-standing metallurgical knowledge. Overall, this work highlights the value of data-driven approaches in materials science: machine learning can complement traditional experiments and simulations by providing rapid, reliable insights into processing structure relationships, guiding heat treatment optimization, and accelerating alloy design. With larger datasets and integration of physics based models, this approach has the potential to support the development of stronger, tougher, and more reliable maraging steels for demanding engineering applications.</p>