Home | Call for Papers | Call for Workshops | Call for Competitions & Datasets
Conference Committee | Contact | Accepted Workshops | Accepted Papers
Competitions & Datasets | Accommodation |  Registration | Keynote Speakers | Schedule

4th ACM International Conference on AI in Finance (ICAIF-23)

Accepted competitions are outlined on this page. The competitions will be highlighted on the second or third day of the conference.

Accepted Competitions

Competition Title	Competition Info
FinRL Contest	See Details
Regime-Switching Financial Time Series Generation	See Details
Trading Hackathon Challenge	See Details
Financial Transaction Matching	See Details

FinRL Contest

Overview

Financial reinforcement learning (FinRL), as an interdisciplinary field of finance and reinforcement learning, has been recognized as a promising approach to financial tasks. Over the past decade, deep reinforcement learning (DRL) has achieved remarkable success in solving complex problems across a variety of domains, including robotics, gaming, and large language models like ChatGPT and GPT-4. The success of DRL has also led to its application in finance, where it has demonstrated great potential for enhancing the performance of financial tasks, such as portfolio management, option pricing, and algorithmic trading.

The FinRL contest is a competition that explores and evaluates the capability of machine learning methods in finance. However, developing machine learning algorithms for financial data presents unique challenges:

1. Dynamic Transitions and Data Noise: Financial datasets are constantly evolving, making it difficult to capture the underlying dynamics. Moreover, financial data is often susceptible to noise and anomalies.
2. Partial Observability: No traders can access complete market-influencing information due to the gigantic, complex, and unpredictable nature of financial systems, where unforeseen events like natural disasters, policy changes, and consumer behavior shifts can have hard-to-anticipate effects.
3. Other Complex Behaviors: Many other factors can cause complicated and unpredictable behaviors, such as the market’s decentralized nature and the large number of financial instruments available.

The FinRL Contest presents two tasks, data-centric stock trading and real time order execution, aiming to foster innovations toward addressing the above challenges. We welcome students, researchers, and engineers who are passionate about finance and machine learning. And we encourage the development of tailored data processing strategies, novel features, and innovative algorithms that can adapt to changing market conditions and generate superior returns for investors.

Organizers

• Keyi Wang, Master’s candidate at Northwestern University
• Ziyi Xia, Master’s candidate at Columbia University.
• Kent Wu, Master’s candidate at Columbia University
• Ethan Havemann, Undergraduate at Northwestern University, Founder of Northwestern Fintech
• Jiale Chen, Undergraduate at Northwestern University and RL lead for Northwestern Fintech

Regime-Switching Financial Time Series Generation

Overview

The limited availability of financial and economic data is a key bottleneck that limits the successful applications of machine learning algorithms in the finance sector. High-fidelity synthetic data generation provides a promising means to augment data that resembles the statistics of real data while preserving data privacy. Popular generative models include Generative adversarial networks (GANs), Variational Autoencoder (VAE) and more recent diffusion models.

Financial time series have unique characteristics such as time heterogeneity, and low signal-noise ratio. In particular, many economic and financial time series exhibit abrupt changes in the patterns, which are often associated with events, such as financial crises, pandemics, and government policy changes. An important illustrative example is the financial price time series, which is typically affected by the fundamentals of markets. The dynamics of the asset price processes behave in significantly different ways during periods of financial crisis compared to more economically stable periods.

Developing robust and high-fidelity generative models is crucial for simulating synthetic financial time series, especially for the financial dataset exhibiting structural shifts. If these models overlook structural changes in patterns of financial time series, it could result in significant financial losses and inaccurate risk assessment. Thus, this competition seeks to engage participants in advancing the generative model methodologies for handling regime-switching time series data.

Organizers

• Hao Ni, UCL, The Alan Turing Institute, DataSig
• Lukasc Szpruch, University of Edinburgh, The Alan Turing Institute
• Jiajie Tao, UCL, DataSig

Trading Hackathon Challenge

Overview

We’re excited that you’re here to participate in our event. Whether you’re a coding novice or an experienced developer, this hackathon is designed to provide you with an engaging and challenging experience.

The objective for you in this hackathon is to build an equity trading algorithm that maximizes profit by buying and selling stocks (i.e., manage an investment portfolio). You will be provided access to years of historical data for a group of 500 stocks. Your algorithm should step through the data, one day at a time, and for each day, generate orders to buy or sell each of the 500 stocks. You can work individually or form a team to work together. Best of luck with your projects! We can’t wait to see what you come up with!

Organizers

• Haibei Zhu, JP Morgan Chase & Company

Financial Transaction Matching

Overview

One of the benefits of applied machine learning is the enhanced automation of tedious tasks currently performed by middle and back office operations team, freeing them up for more creative and productive work.

One such ubiquitous and time consuming back office task in the financial services industry is the matching of financial transactions. This function is performed in all organizations and can include matching and reconciliation of bank statements, trades, financial positions and other financial transactions. 

Currently due to data quality limitations, 100% of financial transactions do not match automatically, leaving transactions to be matched manually.  In the payments field alone, over 17 billion transactions are settled by the Federal Reserve each year. Typical automated match rates for these types of transactions is often less than 90%, leaving 10’s of millions of transactions to be manually processed. Typically reconciliation teams spend 30% of their day on this manual matching leading to delays in the completion of reconciliations.

The competition consists of creating a machine learning model which is able to perform matching between two sets of financial transactions with both a high automated match rate and high accuracy (both of which are the key operational metrics used in reconciliation departments).

Organizers

• Tracey Lall