Sales League Table 2020 | Banking Technology Winners

Results Announced!

Stay up-to-date with the latest industry news and analysis.

Subscribe to BankTech Daily News today. Limited time offer.

The Black Swan Opportunity | Get your bank digital ready.

IBSI Special 5 Digital Report Package with Special Offer. Subscribe now

Core Banking Market Dynamics Report 2020

Global analysis of core banking system sales. Subscribe now

ML algorithms learning investment signals

Machine learning and associated algorithms are making real waves not just in banks’ front offices but also in analysing and spotting trading opportunities in the stock markets.

Machine learning (ML) is being used to identify trading patterns, initially in historical trade and quote data. This may sound familiar to those who are cognisant of technical analysis and the end goal is indeed the same, that is, to find useful patterns in historical and even real time data that lead to decisions that may result in profitable trades either long or short.

According to Tom Finke, head of machine learning product management at software and data provider OneMarketData: “What’s different now is that the techniques have evolved in performing that analysis. In particular, we are able now to use machine learning algorithms to help improve some of the more historical sorts of algorithms that were used to try to detect patterns. When we train machine learning models, such as neural network models, they are able to find patterns that traditional analyses like regression analyses might not otherwise find. These sorts of algorithms are able to find patterns that mere humans are not able to find because of the extent of the vastness of the data that can be analysed.”

An arms race?

Of course, if every trader was to follow the same analytical signals they would all be doing the same thing at the same time. Finke admits that funds, brokers, trading firms, banks and asset managers are facing a “sort of arms race”! He added: If you don’t participate, there is a risk, you’ll indeed be left behind. It would be advisable for investment firms and investment funds that want to stay on top of things that they really should form teams to at least investigate how machine learning can help with their investment and trading decisions.”

But the machines are not completely taking over, well not yet. “It takes some human ingenuity and cleverness, to decide the parameters around those machine learning algorithms. For example, what is the most appropriate data set on which to build a machine learning model?”

Right now, the human element is still required. It takes a person to decide which ML algorithm should be used and what pool of data should be analysed. However, there are companies working on this with ML algorithms being developed with the aim of having them choose which are the best ML algorithms to use.

Watching the markets move

With algorithms being used to analyse trading patterns it should come as no surprise that one way this is being leveraged is in market surveillance. OneMarketData’s core product is a time series tick level database on which that company had built various vertical applications – one of the most popular being trade surveillance.

“We have that particular application being used by a major exchange in the US and by quite a few investment banks. They’re analysing order books; some historically and a few in real time to try to detect patterns that are nefarious such as spoofing  or layering [both forms of illicit manipulation in which a trader may attempt to deceive others regarding the true level of supply/demand for a given financial instrument]. Traditionally there are patterns that you can look for and detect to try to find these activities and now we are in the early stages of applying machine learning algorithms to that,” said Finke.

One thing that has changed in the financial markets in the last few decades is the sheer volume of data and number of trades. Finke noted that in the last week of February 2020 when the world’s financial markets became seriously ‘spooked’ by  concerns over the global Coronavirus (COVID-19) outbreak, the number of ticks during the volatility was such that “some of our customers were running out of memory in their memory databases”.

Seeking the right signal

ML and the appropriate algorithms are capable of analysing more than just price data. For example, Bloomberg now provide a machine-readable news feed that is tagged to make it easier for computer software to parse the text.

This parsed news may then be stored in the same way as trader quote ticks. Run it through a natural language processing algorithm to parse it into meaningful chunks (a technical term!) and then as stage two use another ML algorithm to decide whether there is enough information to provide a trading signal… and if there is, what should that signal be?

Such algorithms are being developed with the analysis of historical data but once the models are trained, they can be applied to real time streaming news data to try to generate real time trading signals. Sure-fire success in such endeavours is by no means guaranteed. “This still a hard problem. It’s hard for humans, it’s hard for anybody to pull meaningful market sentiment out of newsfeeds. We’re still in the early stages of having any sort of effective results, but that’s certainly not stopping people from making the attempt,” concluded Finke.

 

Related IBS Intelligence Research

Related Posts

BNY Mellon to boost real estate fund administration services for Deka Immobilien

BNY Mellon announced today that Deka Immobilien has selected the former’s services for the provision of global fund administration coordination and oversight for its Special Purpose Vehicles (SPVs). BNY Mellon is expected to coordinate accounting data from the SPVs around the world in which Deka funds invest, review the data, translate it into the required […]

This post is only available to members.

Read More »

PPRO study shows younger generations leading the UK alternative payment method adoption

A new study by PPRO has revealed that younger generations are spearheading the adoption of alternative payment methods (APMs), including bank transfers and e-wallets. The study showed that 42% of the Millennials and 35% of the Generation Z respondents have either used the above-mentioned methods of payments before or are confident using them. UK consumers […]

This post is only available to members.

Read More »