Alan Turing is heralded as the father of computing. It is therefore fitting that an institute bearing his name has been established with the aim of helping the UK lead the world in the next phase of harnessing the power of computers in everything from medicine to financial regulation.

The Alan Turing Institute, announced by Chancellor George Osborne in September 2014, seeks to reinforce Britain’s skills in the analysis and application of 'big data'. The buzz phrase roughly translates to the mass of information that is now available, thanks to both the exponential growth in computer power and to our increasing adoption of technology to handle all aspects of business and domestic life. The institute will be formally launched in November this year. A series of academic workshops and data summits with delegates from academia and industry are set to take place that will determine the key areas for research. The financial services sector is likely to feature heavily.

Philip Treleaven, Professor of Computer Science at University College London – one of the five university partners in the Turing Institute – is organising the financial services summit, which will take place on 14 October. “Financial services – investment banks, hedge funds, fund managers and so on – have been doing quantitative work for years,” he says. “It is largely maths driven; using mathematical models, machine learning, modelling behaviours and similar techniques.”  

In hedge funds, for example, the use of algorithms underlies much of the high-frequency trading that aims to exploit price anomalies and other market distortions. However, algorithms and software are becoming ever-more complex as, for example, Knight Capital found out – the trading firm fell victim to a rogue algorithm that ultimately cost the business its existence.

The real timeBrandon Davies, a director of IT firm Obillex, says that high-frequency trading has reached such a point that protagonists are now debating the definition of real time: “If trades have to be carried out in sequence, but across more than one international market, you are dealing with nanoseconds, but atomic clocks in different areas are not necessarily the same. We have got to the situation where we may not even be able to agree what real time is.” Issues like this could, he adds, cause real problems for regulators – their future success will depend on how good their systems are.
“The amount of data available will continue to increase by a factor of ten every few years, rapidly bringing new applications for machine learning”

Regulation is one of the three key areas where Treleaven thinks data science is already helping transform the industry.

1. Recommender systems in retail finance banks are learning from the data-mining techniques used by supermarket loyalty cards, which retailers use to suggest purchases and drive sales. Crowdsourcing companies and peer-to-peer lenders are also using similar systems to help service their customers.

   Lending Club, for example – the fast-growing internet-based US retail finance business – uses metrics like the length of time a borrower has had an email account, the way they interact with the site and their use of social media to help with their credit score. Its data analysis means it can constantly update these metrics to include those which are best at predicting defaults, and can refine them for different geographies and markets.

2. Fraud-detection systems use mathematical techniques to interrogate data, looking for unusual patterns. The London Stock Exchange, for example, has been using such systems for 25 years to help it uncover instances of insider dealing.

3. Mathematical techniques can be employed to collect and analyse data within the fields of regulation and compliance, says Treleaven. That can be through techniques such as data mining and analysis, or through simulation. “Regulators can build a model of a financial market so that they can understand how it will work,” he adds. “Or they can build a model of systemic risk, so that they can understand what will happen if a particular bank gets into trouble.”

The Financial Conduct Authority’s Project Innovate specifically cites “discussions with fintech businesses, trade bodies, academics, consultancies and others to identify potential ways to support the adoption of new technologies to facilitate the delivery of regulatory requirements. The aim is to help firms to develop and benefit from new technology and comply with regulation in a cheaper and easier way.”

The not so newHoward Covington, Chair of the Alan Turing Institute and a former City investment banker, thinks a deeply interesting area for the future is machine learning, where computers use large sets of data essentially to programme themselves. “Machine learning will have a very big impact across the financial services sector,” he says. “Its impact will stretch from developing hedge fund strategies at a very technical level to interactions with customers in a retail environment.”

That is being spurred by technological advances allowing processing power and access to data to grow dramatically. Covington points out that machine learning has been around for at least 20 years. What’s new is the huge increase in the amount of data from which machines can now learn. “Detecting patterns from 100,000 examples proved to be not that effective. Now, systems can learn from ten million examples and suddenly machines have begun to get very good at spotting patterns. The amount of data available will continue to increase by a factor of ten every few years, rapidly bringing new applications for machine learning.”

Another key area for development is blockchain technology, which is behind virtual currencies like Bitcoin and which, Covington says, is attracting “a huge amount of research and thought”. Banks are already investing heavily. In September, for example, nine of the world’s biggest banks – including Goldman Sachs and Barclays – have joined with a US technology firm to explore ways of using blockchain technology in financial markets.

The first of many? Organisations like the Alan Turing Institute are being tried in other countries, including the US, France and Germany. The success of Britain’s institute will largely depend on input – of ideas as well as funding – from the industry as well as academia. The Government has provided £42m over five years, and five university partners have provided £25m, but more will be needed if it is to meet its aim of putting Britain at the forefront of data science. Covington expects half of its effort will be directed at foundational research and half at getting the research into commercial and industrial applications. The institute could be funding “a couple of hundred” postgraduate students within the next few years, he says.

“It is now relatively inexpensive to store and manipulate large amounts of data, particularly because of the advent of cloud computing,” concludes Covington. “The issue is finding the talent – the people with PhDs in data science – to do the work.”