Transforming markets: How AI-driven insights are shaping the future of finance

Image
Cornell Hall

This article originally appeared in the Fall 2024 issue of Trulaske Magazine. 

By ERIC STANN with MU News and Information

Using artificial intelligence (AI) could be a game changer in the fast-moving and constantly changing world of finance helping drive greater efficiency, smarter decision-making and stronger risk management. As markets grow more complex, AI can provide the tools for investment professionals to stay ahead of the curve.

AI aids the analysis of vast amounts of data — both structured (such as financial statements) and unstructured (such as news or social media sentiment) — to extract actionable insights. By using machine learning algorithms, financial firms can make more accurate predictions, identify trends, and optimize investment strategies.

Kuntara Pukthuanthong
Kuntara Pukthuanthong

Kuntara Pukthuanthong, the Robert J. Trulaske, Jr. Professor of Finance, is developing several AI-enabled tools that investors could one day use in their daily work. 

Variational Recurrent Neural Networks (VRNNs)

Much like scenes in a movie, this AI model transforms complex financial information into graph-based visualizations to help predict stock prices. The model's ability to predict pixel changes in market narratives translates into robust forecasts for weekly returns — outperforming traditional price trend strategies while also considering firm-specific characteristics. 

“Financial markets are not static entities; they pulsate with life, evolving and reacting to many stimuli,” Pukthuanthong said. “This dynamism is reminiscent of frames in a cinematic reel, where each frame, though a standalone snapshot, is intrinsically linked to its predecessor, painting a broader narrative.” 

The significance lies in the results: achieving a Sharpe ratio of 2.94 for equally weighted portfolios and 2.47 for value-weighted portfolios demonstrates the model's ability to deliver strong risk-adjusted returns over time. By achieving an alpha of 55 weekly basis points adjusted for risk factors, this approach shows potential to significantly outperform conventional models, making it a powerful tool for investors seeking more precise market predictions.

Assessing firm similarities 

Using AI, Pukthuanthong introduces an innovative method to assess firm similarity using visuals, which has significant implications for financial markets and investment strategies. By analyzing four million images representing companies’ operations, the concept of Image Firm Similarities (IFS) offers a novel approach that is more dynamic and potentially superior to traditional classification systems like SIC, GICS and NAICS.

As businesses rapidly evolve, classification methods that adapt swiftly to firms’ operational focus changes are needed,” Pukthuanthong said. “An innovative clustering approach should offer the flexibility to reflect these rapid shifts and allow companies with diverse activities, such as Tesla, Amazon and Walmart, to simultaneously belong to multiple industries, better capturing modern enterprises’ multifaceted nature.”

IFS mimics how the brain processes visual information, enabling better alignment with investor-defined peer groups. This means that IFS can identify which companies are truly similar in terms of operations, not just based on conventional industry codes or textual analysis. As a result, it performs well in strategies like pair trading (where two similar stocks are traded together), diversification, and identifying industry momentum (tracking the overall trends of industries).

Processing financial information

Pukthuanthong challenges traditional assumptions about how investors process financial information, highlighting the significant role that media distortion plays as news spreads through the market. Rather than attributing investor behavior solely to cognitive biases like overconfidence or the use of outdated information, the study shows that the distortion of news stories before they reach investors can be a key factor in influencing market behavior.

“The transmission of the original story in retelling articles could potentially be biased,” Pukthuanthong said. “Several factors can drive the bias, including memory, social and motivational factors and the specialization of news outlets.”

This research enhances the understanding of how the media ecosystem influences investor behavior and market dynamics and suggests that the financial sector reconsider how financial news is consumed and acted upon in investment strategies.

Pukthuanthong’s papers, “Just Look: Knowing Peers with Image Representation,” “Animating Stock Markets” and “Transmission Bias in Financial News” are being considered for publication.