Machine Learning Techniques for Variable Annuity Valuation

摘要

Machine learning refers to a broad class of computational methods that use experience to improve performance or to make accurate predictions. There are two broad categories of machine learning tasks: supervised learning and unsupervised learning. Supervised learning tasks involve labeled data, which consist of inputs and their desired outputs. Unsupervised learning tasks involve unlabeled data, which consist of only inputs. In this paper, we give a brief overview of some machine learning techniques and demonstrate their applications in insurance. In particular, we apply data clustering and tree-based models to address a computational problem arising from the valuation of variable annuity products. Our numerical results show that tree-based models are able to produce accurate predictions and reduce the computational time significantly.