Artificial Intelligence and Machine Learning for Real-Time Risk Assessment in Group Insurance and Retirement Investments

Abstract

This body of work details a sophisticated investment evaluation service for retirement savings, Roth, SEP, and traditional IRAs, as well as for employees in group 401(k) plans and pension, profit-sharing, and thrift plans. The service utilizes AI and machine learning to forecast the trajectory and risk of returns from the stock and bond markets, while simultaneously projecting the contribution of portfolio performance to participants’ retirement wealth and estate value. This predictive power enables focused analysis of investments subject to correlations with the performance of broad-based market indices and testing of the capability of 401(k) options and an adviser/manager using the forward-looking evaluation of stock/bond probabilities to monitor/adjust recommended and selected portfolios responsibly. The analysis is particularly useful for identifying high-growth low-risk equity investments and both equity and bond alternatives for socially responsible and theme investment portfolios.

Forecasts are generated using a proprietary one-factor ARMA family time series model using machine-learning-enhanced estimation/extrapolation techniques called every-nth-point modeling. The extrapolation technique offers a tractable model with few parameters to estimate that can contain sharp local bumps, runs, and notable declines. The model has been used to forecast a wide variety of economic and financial series—money supply growth, inflation rates, short- and long-rate Treasury yields, stock returns, gold prices, and exchange rates—as well as volatility of individual stocks and stock indices, and interest rates from the term structure.