Abstract: In order to realize the batch prediction of hot-spot temperature（HST） for traction transformer group in the high-speed railway to evaluate their insulation life. The HTS, load factor and ambient temperature measured by a traction transformer A equipped with fiber-optic probes were divided into training set and prediction set. Genetic programming was used to drive training set modeling, and an explicit expression prediction model which could directly evaluate the dynamic HST was established. The longitudinal prediction of the prediction set and the horizontal prediction of transformers B and C under different sections were carried out for the THS. The study results show that the HST is positively correlated with the traction load factor square and the structure of the model is simple and intuitive. The mean relative error（MRE） and the goodness of fit（R2） of the HST prediction under the prediction set are 1.63% and 0.9830 respectively indicating the model has high prediction accuracy and exactitude. The THS of transformer B and C can be predicted well for the MRE is 1.00% and 1.89% respectively. The model has strong generalization performance and the HTS of transformer group under the same service railway line as traction transformer A can be predicted in batches by it.