Proxy models are typically used when calculations need to be performed significantly faster, often by several orders of magnitude, than can be achieved using traditional actuarial models. In these cases, simplified models offer improvements in speed, at the cost of an acceptable level of approximation on the results. The Prophet Proxy Fitting Library is capable of performing both replicating portfolio and curve fitting techniques, including Least Squares Monte Carlo.
The Prophet Proxy Fitting Library uses Replicating Portfolio techniques to find portfolios of assets that closely match the cashflows from a set of liabilities. The asset and liability cashflows used can either be created within Prophet or from external sources. The library is very flexible, allowing multiple constraints including non-linear and custom constraints.
The Prophet Proxy Fitting Library uses curve fitting techniques, including the Least Squares Monte Carlo approach, to generate a polynomial function of risk factors whose result closely matches results from an explicit actuarial model of your insurance business. The library offers flexibility around the type of polynomial to fit, and a choice of regression techniques including the efficient stepwise regression algorithm to identify statistically significant polynomial terms.
The code base is a Prophet library, so it can be integrated into your Prophet modelling process, and it meets the same standards of transparency, flexibility and robustness as the other Prophet libraries. The library is accompanied by a user reference, Quick Start Guide tutorials to create working examples of both replicating portfolios and risk factor polynomials, and Excel-based analysis tools incorporating risk factor analytics capabilities.
For more details, please see the Prophet Proxy Fitting Library Datasheet.