AI-based Robo-Advisor for Investments

The Robo-Advisor based on Reinforcement Learning promises optimized and adaptive automated investment decisions.

IntelliInvest RL: Introduction and Features

IntelliInvest RL is a project that originated as an internal Proof of Concept (POC) and was later presented at the “TechFS AI Challenge – 2024”. It aims, in the context of the evolution of digital financial services, to provide a robo-advisory solution designed to offer automated financial advice. Based on Reinforcement Learning models, the system can analyze an investor's risk profile and deliver a personalized investment portfolio tailored to their needs, constantly adapting to market fluctuations.

Workflow

The process begins with the completion of a risk questionnaire compliant with MiFID regulations, the data of which is engineered to be used by the robo-advisor's AI system.

Financial updates are retrieved daily from sources like Il Sole 24 Ore or, more generally, from selected RSS aggregator sites. Sentiment analysis is then applied to analyze the news to assess the impact of the information on the portfolio's assets. Sentiment analysis is a technique used to understand the emotions or opinions expressed in a text. Essentially, it analyzes sentences or words to determine whether the message has a positive, negative, or neutral tone.

Based on the risk profile, the AI model processes the investment portfolio and adjusts its values according to the daily influences of the financial market, ensuring proactive and personalized investment management.

System Architecture


The system supports three different risk portfolios: low, medium, and high, which include a combination of assets such as bonds (e.g., Amundi US Treasury Bond Long Dated UCITS ETF and iShares Core UK Gilts UCITS ETF) and stocks (e.g., Amazon, NVIDIA, and Alphabet).

The core of IntelliInvest RL is based on the FinRL architecture, an open-source platform for automated trading through Reinforcement Learning. This approach allows the AI agent to learn from its actions, continuously refining investment decisions to maximize cumulative profit.

Finally, sentiment analysis is performed using a Hugging Face BERT language model, dedicated to natural language analysis, based on the Transformer architecture.

Conclusion

IntelliInvest RL represents a useful and practical experiment in the field of automated financial advisory. By utilizing advanced technologies like Reinforcement Learning and sentiment analysis, it offers a modern approach to portfolio management.

This solution not only meets the immediate needs of investors but continuously adapts to market conditions, making it an essential tool for those looking to optimize their investments in a rapidly evolving financial context.