The Department of Engineering at Aarhus University is heading a new project to put an end to stock-exchange fraud and market manipulation. The researchers will use artificial intelligence (AI) to clean up the extensive fraud taking place, where control is currently implemented via manual sampling.
A team of researchers from Aarhus University (AU) has just received a grant of DKK 2.8 million from the Independent Research Fund Denmark for a project that may impact share trading throughout the world.
The team is being headed by Alexandros Iosifidis, an associate professor at the Department of Engineering and an expert in machine learning. The project aims to develop a system that can identify suspicious trading on all of the world's stock exchanges - something that is currently being done through manual samples.
"The AI solutions we're looking for require far less manual work, they reduce costs, and they're much more effective than today's controls," says Alexandros Iosifidis.
As yet, no one knows the extent of stock-market fraud globally, but the US Federal Trade Commission (FTC), has estimated that, in the US alone, the problem amounts to between USD 10-40 billion a year.
There are many different types of stock-market fraud, which is a ticking bomb under the whole philosophy behind the market-economy principle of supply and demand.
"A reliable, automatic safeguard against market manipulation could therefore be crucial for transparent stock markets," says Alexandros Iosifidis.
The project will design entirely new methods which, using machine learning, will be able to conduct systematic analyses of stock-exchange activity, recognise irregular transactions, and identify patterns in share-price data. This will make it possible to detect, and even predict, market manipulation in share trading across the entire world.
"We're working on the assumption that trading in one share affects future trading in other shares in specific patterns that can be recognised using data-driven analyses. We're focusing on recording trading activity based on jumps in average share prices, and we will detect these in actual stock-exchange data in order to identify relationships between share transactions and irregular trading activities," he says.
The project is called 'Data-driven Inter-Stock Predictive Analytics' or 'DISPA', and it will run for three years, coordinated by Associate Professor Alexandros Iosifidis.