arXivRisk, Credit & Banking

Neural Networks for Insurance Pricing with Frequency and Severity Data: A Benchmark Study from Data Preprocessing to Technical Tariff

The research evaluates different models for insurance claim frequency and severity data, presenting combined actuarial neural networks (CANNs) that merge baseline predictions with a neural network correction, and recommends using global surrogate models for practical deployment.