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

AI-Powered Cryptocurrency Portfolio Management System

Researchers from the University of Tsukuba in Japan have recently developed an AI-powered cryptocurrency portfolio management system, dubbed CryptoRLPM, which is the first of its kind to use on-chain data for training.

CryptoRLPM utilizes a technique called “reinforcement learning” (RL) to optimize its model and incorporate on-chain data. RL is an AI system that interacts with its environment and updates its model based on the reward signals it receives.

The architecture of CryptoRLPM is composed of five main units that work in tandem to process and manage structured portfolios. Feedback from reinforcement learning is applied throughout the system.

AI-Powered CryptoRLPM

C3.ai and Argo AI have developed CryptoRLPM, a portfolio management system powered by artificial intelligence (AI). It is comprised of five modules: Data Feed Unit, Data Refinement Unit, Portfolio Agent Unit, Live Trading Unit, and an Agent Updating Unit.

To evaluate the success of CryptoRLPM, the scientists created three portfolios, each with different combinations of Bitcoin (BTC), Storj (STORJ), Bluzelle (BLZ), and Chainlink (LINK). The experiments ran from October 2020 to September 2022 and were divided into three phases: training, validation, and backtesting.

The team then measured the performance of CryptoRLPM against a baseline market evaluation using three metrics: “accumulated rate of return” (AAR), “daily rate of return” (DRR), and “Sortino ratio” (SR).

AAR and DRR are two metrics that give an overview of how much an asset has increased or decreased in a given time period, while SR evaluates the asset’s return in terms of risk.

As per the findings of the recent pre-print research paper, CryptoRLPM shows a substantial improvement in comparison to the existing baseline performance:

Categorized in:

Tagged in: