The Markowitz portfolio framework is extensively used to find out static asset weights, whereas Merton’s dynamic strategy permits allocations to regulate with altering market circumstances however is mathematically difficult and fewer sensible. We deal with this hole by making use of machine studying to dynamic portfolio optimization within the spirit of Merton, incorporating financial regimes outlined by the VIX volatility index. A synthetic neural community is educated to study optimum allocation insurance policies throughout regime-switching environments and is in contrast with classical regime-agnostic and theoretical regime-switching Merton methods. On artificial information with practical constraints prohibiting borrowing and brief promoting, the machine studying technique outperforms conventional benchmarks. Two empirical backtests—utilizing month-to-month information from 1990 to 2025 and annual information from 1928 to 2025—present that accounting for regimes enhances efficiency and robustness.
