Christophe Dutang and his co-authors have been awarded the 2025 Best Paper Award by the journal Variance

Christophe Dutang (ASAR team) and his co-authors have been awarded the 2025 Best Paper Award by the journal Variance, published by the Casualty Actuarial Society, for their work on the contribution of transfer learning to predicting individual premiums. The paper compares learning methods (boosting and neural networks) calibrated with or without transfer against the benchmark method (the hierarchical Buhlmann-Straub method) using non-life insurance data. The results show that transfer learning significantly improves the accuracy of predictions compared with a model trained only on the insurer’s data and provide competitive results compared with those of hierarchical credibility models.