Adaptive Predictive Portfolio Management Agent

The paper presents an advanced version of adaptive market-making agent capable performing experiential learning, exploiting a “try and fail” approach relying on swarm subordinate agents executed in virtual environment to determine optimal strategies. problem is treated as “Narrow AGI” with the scope goals environments bound financial markets, specifically crypto-markets. Such called “adaptive multi-strategy agent” it executes multiple strategies virtually selects only few for real execution. presented extended solve portfolio optimization re-balancing across assets so active management being addressed. Also, attempt made apply learning multi-agent simulation backtesting based historical market data, can learn mappings between specific conditions corresponding these conditions. Additionally, equipped capacity predict price movements social media which increases its performance.