Non-Parametric Integral Estimation Using Data Clustering in Stochastic Dynamic Programming : An Introduction Using Lifetime Financial Modelling

This paper considers an alternative way of structuring stochastic variables in a dynamic programming framework where the model structure dictates that numerical methods of solution are necessary. Rather than estimating integrals within a Bellman equation using quadrature nodes, we use nodes directly from underlying data. An example of the application of this approach is presented using individual lifetime financial modelling. The results show significant benefits to the data-driven approach compared to quadrature and Quasi-Monte Carlo approaches, which hold for both a single stochastic variable and multiple stochastic variables