Researchers from the University of Tsukuba in Japan have created a groundbreaking AI-powered cryptocurrency portfolio management system called CryptoRLPM (Cryptocurrency reinforcement learning portfolio manager). This system is unique in its utilization of on-chain data for training, making it the first of its kind according to the scientists. The AI system employs reinforcement learning (RL), an optimization paradigm where the AI interacts with its environment (in this case, a cryptocurrency portfolio) and adjusts its training based on reward signals.
CryptoRLPM consists of five primary units that collaborate to process information and manage structured portfolios. These units include a Data Feed Unit, Data Refinement Unit, Portfolio Agent Unit, Live Trading Unit, and an Agent Updating Unit. By applying feedback from RL throughout its architecture, CryptoRLPM enhances its decision-making capabilities.
To evaluate the effectiveness of CryptoRLPM, the researchers assigned it three different portfolios. The first portfolio comprised only Bitcoin (BTC) and Storj (STORJ), the second included BTC, STORJ, and Bluzelle (BLZ), and the third contained all three assets along with Chainlink (LINK). The experiments spanned from October 2020 to September 2022 and encompassed three distinct phases: training, validation, and backtesting.
The researchers compared CryptoRLPM’s performance against a baseline evaluation of standard market performance using three metrics: “accumulated rate of return” (AAR), “daily rate of return” (DRR), and “Sortino ratio” (SR). AAR and DRR provide a quick overview of an asset’s gains or losses within a given timeframe, while SR measures the risk-adjusted return of an asset.
According to the scientists’ pre-print research paper, CryptoRLPM exhibited significant improvements over the baseline performance. The AI system demonstrated at least an 83.14% improvement in AAR, at least a 0.5603% improvement in DRR, and at least a 2.1767 improvement in SR when compared to the baseline Bitcoin.
This innovative approach to crypto portfolio management showcases the potential of reinforcement learning and on-chain data for optimizing investment strategies. By leveraging AI-driven systems like CryptoRLPM, investors can potentially enhance their decision-making processes and improve portfolio performance in the dynamic world of cryptocurrencies.