Scientists created a crypto portfolio management AI trained with on-chain data

A study has revealed that a newly developed AI-run cryptocurrency portfolio management system, called CryptoRLPM (Cryptocurrency reinforcement learning portfolio manager), may lead the digital asset space to a new level of operation efficiency. Generated by two experts from the University of Tsukuba in Japan, CryptoRLPM is primarily constructed based on a cutting-edge technique in optimization trajectory – Reinforcement Learning, allowing it to learn from raw on-chain data to shape tailored portfolios that are diversely subclassified into five major categories: Data Feed Unit, Data Refinement Unit, Portfolio Agent Unit, Live Trading Unit as well as Agent Updating Unit. Presenting a practical method to exceed prior benchmarks, CryptoRLPM has been evaluated through marking three portfolios during the period extended from October of 2020 to September 2022 without overlooking any of three important invariant scales: “accumulated rate of return” (AAR), “daily rate of return” (DRR) and finally “Sortino ratio” (SR). Its capacity was conclusively checked off against two distinctive stages in an effort to reaffirm the intended supplement of quality in the fields of digital investments management, resulting in gigantic enhancements of consistent returns for its investors than the traditional market models on every level.

Robert Wilson author
Articles: 12200