AIAF Group

The AIAF Group is a research group at the Instituto Tecnológico de Aeronáutica (ITA) that applies Artificial Intelligence to problems in Finance. The financial domain is especially hard for autonomous systems. Even so, several technologies look promising, and autonomous approaches have some less obvious advantages. Many financial institutions already use intelligent systems in their decision-making and to trade real portfolios. However, building autonomous financial agents that consistently outperform the best human experts remains a distant goal.

AI has many potential applications in Finance, and financial forecasting is among the most studied. Recent work suggests that deep reinforcement learning, Transformers, Large Language Models (LLMs) and related architectures perform well, particularly when combined in ensemble methods.

The digitization of finance has also created new markets, most notably cryptocurrency markets, which include Bitcoin, Ethereum and many other crypto-assets. These markets are highly volatile, and there is no agreement on whether they are viable in the long term or how they should be valued. These features, together with recent advances in AI such as deep learning, open significant opportunities for research and innovation. Financial environments also contain many pitfalls, however, and AI researchers and practitioners working in the area often misunderstand or overlook them.

Deploying autonomous financial agents requires trust to be built into their development, deployment and operation. Trustworthy artificial intelligence is commonly based on five principles: (1) beneficence, (2) non-maleficence, (3) autonomy, (4) justice and (5) explicability, as proposed by Floridi and Cowls in the 2019 paper, A Unified Framework of Five Principles for AI in Society. Accountability is arguably closely tied to the first two principles and to the reasonable management of conflicts of interest. Conflicts of interest among analysts, managers and investors are well documented in finance. Software verification and validation can constrain an autonomous agent’s objectives, or at least make them formally verifiable, so such conflicts can be avoided or at least managed more effectively than with human agents.

The coordinator of the AIAF Group, Prof. Paulo André L. Castro, has published the e-book Introduction to Artificial Intelligence in Finance, which is intended for researchers starting out in the field. The e-book is available for free download!

Introduction to Artificial Intelligence in Finance. (e-book)

Related Conferences and Journals

Scientific Journals

Computational Economics Journal

ACM Transactions on Economics and Computation

Applied Intelligence

Journal of Computational Science

Financial Innovation

The Journal of Financial Data Science

Evolving Systems

International Journal of Finance and Economics

International Journal of Data Science and Analytics

Quantitative Finance

Expert Systems with Applications

International Journal of Intelligent Engineering and Systems

Journal of Autonomous Intelligence

Workshops and Conferences

ACM International Conference on AI in Finance

Advances in Financial Technologies - AFT

WINE - Conference on Web and Internet Economics

ACM Conference on Economics and Computation (SIGECOM)

International Conference on Enterprise Infomration Systems (ICEIS)

Future Technologies Conference

IEEE/WIC International Conference on Web Intelligence and Intelligent Agent Technology

IEEE Conference on Artificial Intelligence (IEEE CAI 2024)

Brazilian Conference on Intelligent Systems (BRACIS)

Encontro Nacional de Inteligência Artificial e Computacional (ENIAC)

Workshop of Artificial Intelligence Applied to Finance - WAIAF (Organized by AIAF Group)

AIAF Publications

Current Projects

Project Title: Artificial Intelligence in Digital Finance and Cryptomarkets: Challenges and Opportunities

Project Funded by FAPESP.

Start Date: July, 2022, End: August, 2024

Coordinator: Paulo André L. Castro. Members: Attilio Sbrana, Fidel Esteves, Gustavo de Freitas Fonseca, Ricardo Gabriel Pontes Lins

Abstract.Artificial Intelligence (AI) and its related fields such as Machine learning and Robotics are changing the world, as we know it. It is also true for the financial field. Furthermore, the progress of digitization in finance has created new markets, specially the so called cryptomarkets, which includes Bitcoin, Ethereum and many other cryptoassets, which are characterized by high volatility and different views about their viability and methods of valuation. These features and breakthroughs in AI, such as deep learning, bring opportunities for innovation and research. However, there are also many pitfalls in financial environments, which are often misunderstood or simply unknown to many AI practitioners and researchers addressing it. Financial problems may be much harder than recognizing faces or driving cars, because patterns often evolve as time goes by and which information is relevant may change with the financial assets. In this initiative, we intend to (1) clarify and describe the financial environment for AI researchers and practitioners including the several pitfalls (2) develop conceptual and methodological procedures in order to tackle these problems in both traditional and cryptomarkets (3) identify or create tools to support the application of the defined procedures. This project benefits from recent and relevant work published in the intersection of Economics and Computation. In fact, ACM created a special interest group to encourage research at the interface between economics and computation, including Finance, Artificial Intelligence and many other fields. Initiatives as this one can benefit both areas and may have impact in practical applications not just, as AI based solution to financial problems, but it may also contribute to development or improvement of new AI algorithms

ProjectTitle: Data Analysis of the Clear Challenge dataset including Questionnaires and Trading Behaviour

Project Funded by XP Investimentos

Start Date: November, 2022, End: August, 2023

Coordinators: Paulo André L. Castro, Prof. Dr.; Carlos Forster, Prof. Dr.

Abstract.The project consists of analyzing data collected during the Clear Challenge to assess the impact of this competition on the risk behavior of day trader operators. The Clear Challenge is a competition sponsored by Clear Corretora and Grupo XP that aims to aims to educate for risk management in Day Trade in a practical way, encouraging effective change in customer behavior. In this project, we proposed a methodology for the analysis of questionnaire data along with its application on discovering insights from investor data motivated by a day trading competition. The questionnaire includes categorical questions, which are reduced to binary questions, yes-or-no. The methodology reduces dimensionality by grouping questions and participants with similar responses using clustering analysis. Rule discovery was performed by using a conversion rate metric. Innovative visual representations were proposed to validate the cluster analysis and the relation discovery between questions. When crossing with financial data, additional insights were revealed related to the recognized clusters.