قالب وردپرس درنا توس
Home / Science / Banks look to the stars to spot trading mistakes

Banks look to the stars to spot trading mistakes



Machine learning technology, which protects Europe's space missions, could soon be used to reduce the risk of "fat fingers" after the European Space Agency has agreed its first collaboration with the financial services industry.

ESA collaborates with financial analyst Mosaic Smart Data to find other applications for algorithms that ensure the safe operation of satellites by monitoring and analyzing their tens of thousands of instruments for early signs of problems.

Mosaic Smart Data hopes that the same underlying algorithms can be used to monitor and analyze millions of data points in the financial market, so that errors and fraud can be detected earlier.

"We have an incredible opportunity to see some of the most advanced data analysis models in the world for our clients' financial market issues," said Matthew Hodgson, Chief Executive of Mosaic Smart Data.

The collaboration was agreed under the ESA Business Applications program, which provides companies with access to ESA technology.

Mosaic Smart Data works with four of the largest banks in the world; Mr. Hodgson said that her first feedback is that "this is exactly the kind of technology that they think will have value." [259] Few finger trades where a trader accidentally presses the wrong key have been a persistent problem for banks around the world, most recently due to a case in South Korea involving 2.8 billion Samsung shares valued at about $ 1

00 billion $ were accidentally spent on employees.

Other prominent cases include the inadvertent payment of $ 6 billion to a hedge fund customer who later repaid the money. ESA technology could also be used to detect other trade anomalies such as fraud.

Banks already have a variety of systems and analytics to minimize the risks of bad transactions. Mr. Hodgson said the ESA algorithms have the potential to do more.

"Banks rely on data from past crimes [to spot outliers]," he said. "There are too few of these cases to create highly robust models and such models, and such analyzes are often based on rather inflexible rule-based approaches, such as when an alarm is triggered in the event of a break."

The ESA models, on the other hand, are based on monitoring current market activity and can be used to analyze banks' internal data, third-party market and economic data, and other factors. "These models are much more subtle," said Mr. Hodgson, "able to detect out-of-norm behavior even if it's within a certain threshold."

A team of four data scientists at Mosaic Smart Data will try to adapt the ESA algorithms for the financial services sector in the coming months. Mr. Hodgson said he hopes a product will be ready for the market by the third quarter of 2018.

In order to be useful in the financial markets, the algorithms must be able to analyze millions of inputs; much more than the "tens of thousands" in ESA's satellite program. Mr. Hodgson said that increasing technology is not a problem and that the biggest challenges lie elsewhere.

"Data from satellites is fairly regular, but financial market data is far less predictable and flowing," he said. "Adapting the algorithms for these more complex data is the key challenge we need to address in this project.

" However, we are confident that this project will produce useful technologies for the financial markets. We think that the use case for these novelty detection algorithms is very clear. "

Mosaic Smart Data does not have to pay the ESA a fee for the development of the algorithms, but must share the results of their research, so the improved algorithms can also be used by the ESA.


Source link