Over the past two years, banking, investment and financial institutions have shown significant interest in machine learning (ML) possibilities, specifically regarding quantitative finance.

In late October 2022, the African Institute of Financial Markets and Risk Management (AIFMRM) hosted a masterclass on machine learning for quantitative finance in Johannesburg, South Africa. It was an event for specialist finance professionals, academics, and researchers. What made this event especially interesting was the high number of attendees. While 55 people may not sound like a crowd, the numbers were significantly larger than similar events in Europe, for instance, where an audience of 15 or 20 people could be expected.

Apart from the numbers – it is also worth noting the high level of interaction, which could be seen in the way participants asked questions, engaged in discussions, and asked about follow-up conversations and possible future collaborations. More companies enquired about application methods, implementation possibilities, and specifics in the research compared with a related masterclass presented two years ago.

This masterclass, aimed at the financial services industry, presented the latest research on the generation of synthetic market data and how machine learning techniques can extract characteristics and qualities to help identify more risks and expose new possibilities – which may be especially valuable for commodity traders and those in derivatives pricing, for instance. Synthetic data generation helps identify specific scenarios and risks that may not have arisen with existing models.

It also lays the basis for applying other techniques of machine learning based methods. Such methods depart from standard assumptions like governing stochastic differential equations. Techniques like Deep Hedging or Reinforcement Learning can be applied to more general set-ups.

It can be challenging to explain in layman’s terms the value of synthetic data generation and machine learning techniques. One way of looking at its relevance is to take something like the oil price. Oil is one of the world’s most important commodities, contributes to a third of global energy consumption, and is used in products ranging from transportation to plastics. Crude oil price fluctuations significantly impact global economies, as Shanu Jain and Ajay Gupta explain in an article on forecasting crude oil prices. Price forecasting is essential to a wide range of stakeholder Interest in financial event shows increasing global appetite for machine learning (ML) ders and can help reduce risks related to oil price volatility.

Another area where these methodologies can help predict certain outcomes with greater accuracy is in determining electricity prices – a crucial sector with massive repercussions and impact, not only in South Africa but worldwide, when considering how the tensions in Russia and Ukraine have affected the energy situation in Europe.

For example, in a research paper published in 2021, Stanford University’s Dr Nathan Ratledge and others demonstrate how sparse data on key economic outcomes limit the development and efficacy of public policy. Their paper goes on to show that this challenge can be addressed by using satellite imagery and machine learning to evaluate the electrical grid in Uganda and find local measurements for the impact of electricity access on livelihoods. They were able to calculate how much grid access improved the asset wealth of villages, helping to provide much-needed data on the impact of grid-based infrastructure investments.

Over the past two years, we have seen greater adoption of virtual platforms and digital solutions in various industries as more organisations embrace remote working conditions and artificial intelligence (AI) software technology to speed up processes and improve work outcomes.

However, we have to outline that the presented methods are grounded in statistical analysis and do not relate to what is commonly known as AI. It is more like recognising and generating patterns observed in data. The underlying relationships of the used data and the features are often non-linear, and the methods help to understand this relationship.

It has brought increased comfort around new technology, like ML and AI. In the coming years, there will be even greater adoption and application of new techniques like synthetic data generation. It brings faster processing and greater depth and efficacy to financial and trading modelling and insights.

Additionally, the AIFMRM masterclass also looked at new research published in the quantitative finance field this year around Stochastic Local Volatility and specifically data-driven and model-free approaches for computing conditional expectations. The new method combines Gaussian Mean Mixture models with classic analytic techniques based on the properties of the Gaussian distribution. The importance of this work lies in the technique’s ability to produce accurate estimates and hedging strategies related to discrete minimal variance hedges.

New methods and techniques employing machine learning are being used more frequently nowadays. A recent financial report indicates a surge in the use of software and new technology in banks and financial institutions globally. The report reveals how advances in machine learning, cloud computing technologies, and artificial intelligence are shaping financial and banking services. Some say that technology and quants (quantitative analysts) are the future of finance. According to Dr Anthony Ledford, chief scientist at Man AHL, a quant-based subsidiary of the Mag Group, which partners with the University of Oxford, the cutting edge of being a quant in finance today is machine learning.

Whether or not one agrees with these statements, machine learning certainly holds much promise for various industries. Increasingly, more avenues are opening for those researching how machine learning can best improve processes and systems in their specific field or industry. It is undoubtedly an exciting time for those exploring the potential of machine learning in quantitative finance.

Professor Jörg Kienitz is an Assistant Professor at the University of Wuppertal in Germany and an Adjunct Associate Professor at the University of Cape Town’s African Institute for Financial Markets and Risk Management