Community Research and Development Information Service - CORDIS

FP7

QualiMaster Result In Brief

Project ID: 619525
Funded under: FP7-ICT
Country: Germany

Automated real-time analysis helps prevent build-up of risky market situations

EU-funded computer scientists have developed an IT infrastructure to monitor risky financial market activity in real time and detect early signals for market failure.
Automated real-time analysis helps prevent build-up of risky market situations
Another global financial crisis could be difficult to prevent without the means to swiftly analyse risks, respond to events and minimise the effect on global markets. Now EU-funded researchers and computer scientists have developed an innovative IT infrastructure to help market analysts, traders and financial regulators to react quickly to potentially risky and unusual market behaviours.

The three-year €2.9 million QUALIMASTER project has developed an integrated system of computer algorithms, models and processes to analyse high-volume financial data in real time to better predict the risk and likelihood of market failures and trigger faster reactions and adjustments.

‘The idea is to improve processing of data streams and big data to build an infrastructure that can adapt very dynamically to changing situations,’ says project coordinator Claudia Niederée, senior researcher at the L3S Research Centre at Leibniz University, Hannover.

Using high-volume data streams from the stock market – sometimes 100 million messages a second – including stock price movements, exchange rates, futures indices and bond prices, as well as social media data from microblogging platform Twitter, combined with information from newswires, financial blogs, governmental data and other sources, the system continuously analyses the situation to detect signs of market imbalance.

‘We investigated correlations to look at systemic risk and to understand the change not only for one individual trade, but also to better understand why markets move and how they move; and also to look into a deeper type of correlation, to see if something happens, how would this pull through,’ says Dr Niederée. ‘We also built visualisations, so that someone who analyses it gets more insights into what to do.’

Dynamic system

All this is done within a fast-moving and adaptive Big Data environment. ‘Existing data-processing systems can be somewhat static, if a sudden surge of financial data due to any kind of crisis occurs there is a time lag to react,’ says Dr Niederée. ‘You would normally have to stop the system, reconfigure it and restart. Our system can do this dynamically and also detect by itself if there is a need for any change.’

‘Like a time machine, it can go also go back to another point in time where the situation was similar to see what happened and to understand the relationships between market players by looking at how they evolve over time,’ says Dr Niederée.

The configurable system, developed in collaboration with financial technology companies, analysed social media using sentiment analysis to see how opinion is formed about certain stocks or companies at any given time; and events detection using Twitter streams to identify signals that affect a certain stock or group of stocks.

Automated hardware reconfiguring

A third component of the project, new in this field, was to translate stream-processing algorithms into hardware code, so that reconfigurable hardware can be dynamically exploited to better cope with unexpected spikes in data activity.

‘If you go into thousands of market players and have to compute millions of pairs of correlations every second, such heavy data loads can be handled more efficiently in hardware,’ Dr Niederée explains.

QUALIMASTER’s work has also attracted the attention of the automotive industry interested in dynamic real-time adaptations to constantly changing situations for driverless cars.

Keywords

QUALIMASTER, finance, fintech, automated trading, market behaviour, stock market, ict, systemic risk, big data, data processing
Record Number: 198816 / Last updated on: 2017-06-07
Follow us on: RSS Facebook Twitter YouTube Managed by the EU Publications Office Top