Professor Juho Kanniainen’s group develops leading machine learning models for financial markets

“We develop extremely fast machine learning models for an exceptionally demanding environment. In principle, nothing should be predictable in financial markets – except at the very shortest time scales, the millisecond level. This is where we focus our efforts.” 

Research combines theory and practice 

Professor Juho Kanniainen leads a research group currently comprising four postdoctoral researchers and a couple of doctoral candidates.  

“Everyone is talking about artificial intelligence, but at its core lies machine learning: the algorithms and models that power this intelligence. Universities have a key role in developing these methods,” notes Kanniainen. 

The group works closely with Alexandros Iosifidis, who joined Tampere University as a Professor of Machine Learning in the spring of 2025.  

“It is fantastic to have Alexandros here. Tampere University brings together world-class expertise in theoretical machine learning and applied data science to explore the dynamics of financial markets. Alexandros’s expertise lies at the core of machine learning, while mine is more on the applied machine learning side. By collaborating, we can achieve results that neither of us could accomplish alone.” 

Researchers in Tampere have been engaged in international collaborations for nearly a decade, helping to position local research at the forefront of this field globally. 

Predicting market moves in milliseconds 

Professor Kanniainen’s best-known research area focuses on limit order book data: the stream of buy and sell orders, their modifications and executions, and the speed at which all this takes place. 

“An enormous number of messages are transmitted within stock markets in mere nanoseconds. Trading algorithms react to each other’s actions in fractions of a second, triggering chain reactions. Genuine predictability only emerges when we focus on sufficiently short time scales.” 

Kanniainen illustrates the extreme nature of this phenomenon with an example from the United States: 

“In 2010, a new cable was laid between Chicago and New York to shave four milliseconds off communication time between these cities. The cost was $300 million. Our perspective is different: instead of speed, we aim for predictability.” 

To achieve this, the group has developed many of the world’s most accurate machine learning models for predicting limit order book data in collaboration with international partners. These models can forecast, for example, the direction of price movements only moments in advance, yet this brief window is enough to make a difference in automated trading environments. 

“Our research has both societal and practical significance. Our methods help to reduce the risks associated with market-making, which improves market liquidity. In addition, our publicly available research results benefit, in particular, small and medium-sized trading firms that are competing against large hedge funds.” 

Studying the spread of insider information through social networks 

Another major research focus for Kanniainen’s group is how information spreads in the stock market through social networks. The group focuses especially on the connections between corporate insiders. 

“Board memberships create a surprisingly dense social network. Previous studies have found that insider information can indeed be transmitted across these networks.” 

The researchers have access to an exceptionally extensive Finnish dataset that allows tracking investors' transactions across all securities, even those outside insider trading regulations.