The glass transition temperature (𝑇 g ) is a parameter used in many glass melt viscosity models as it denotes a temperature around which liquid-glass transition occurs. In this work, 𝑇 g values were measured for a series of low-activity waste (LAW) glasses using differential scanning calorimetry. These data were combined with 𝑇 g data of other waste glasses available from literature. The combined dataset, consisting of 137 data points, was used for the development of several models to estimate 𝑇 g from glass composition. When testing the number of influential components and different supervised learning methods, we demonstrated that using more than 10 components or using non-linear methods brought marginal improvement to the model accuracy. The best model predicts 𝑇 g of both LAW and high-level waste glasses with reasonable accuracy.