4-Xtra Technologies

Predicting extreme events and synthetic data generation utilising AI and machine learning


Introducing the 4-Xtra Platform

Building upon initial academic work in the theory and predictability of extreme events and utilising innovative machine learning and artificial intelligence methods, 4-Xtra has developed a SaaS based extreme event forecasting risk modelling solution and a complementary synthetic data generation tool for the increasingly complex world of the financial markets and beyond. Data and signals can be easily output through the web-based user interface or delivered through our API for integration into your own systems as required. Our solution is applicable to many use cases in financial services such as fraud and credit delinquency forecasting, trading signal generation, portfolio optimisation and ESG related forecasting.

4-Xtra Extreme Forecasting Engine

4-Xtra Extreme Forecasting EngineSaaS based, machine learning risk management platform that is capable of accurately predicting the likelihood of a multi-variate extreme event occurring, together with the associated magnitude of that predicted event using nonstationary time-series data.

The platform allows for any number of adaptable business configurations to be overlaid to manage specific industry risks and produces outstanding results, 6-7 times more accurate and faster than classical risk models based on stationary data.

Leveraging different data sources as inputs such as text, images, tabular and time-series data, the engine can efficiently extract extreme value information from the underlying inputs and make future predictions about upcoming extreme events.

4-Xtra Extreme Forecasting EngineApplicability of the 4-Xtra Extreme Forecasting Engine is extremely wide ranging in financial services alone, with proven use cases such as:

Trading – Forecasting individual stocks, indexes, alternative assets and currency movements
Portfolio Management – Optimisation and improved performance
Lending – Forecasting events to enhance and improve credit decision making and repayment delinquency rates
Insurance – Predicting probability of claims to enhance risk pricing

4-XTRA SYNTHETIC DATA GENERATORProduces unlimited fully artificial data sets which inherit the statistical features of the original input dataset.

Based on a user-uploaded tabular (time-series) dataset (either through our user interface or an API call), the 4-Xtra Synthetic Data Generator is able to generate completely synthetic data that effectively preserves the statistical properties of the original dataset ensuring that data privacy and utility of the synthetic data are maximized.

Furthermore, leveraging the 4-Xtra EV engine within the SDG, the generator can efficiently, on-demand generate extreme samples of the data that go far beyond what was observed in the original dataset. This empowers users to efficiently manage risk by simulating and exploring extreme samples of data that would usually require years to collect.

4-XTRA SYNTHETIC DATA GENERATORApplicability of our synthetic data generator to financial services is vast and can be used in a variety of cases, for example where:

Data is required for testing purposes but is scarce or lacking
Data privacy is required
Regulatory constraints exist on the utilisation of data (location, anonymity requirements) The risk characteristics of an investment portfolio, trading strategy, require testing without criticism of over-fitting models to historical data
Scenario analysis can be performed – data sets can be constructed to test trading models performance against extreme events
Insurance pricing models can be optimised

Who we are

The leadership Team of 4-Xtra Technologies

Our team is composed of seasoned experts in their respective fields.
Learn more about their individual contributions and their collective efforts in making this innovative breakthrough possible.

David Potter - CEO

David Potter


Colin Day - Non-executive chair

Colin Day

Non-Executive Chair

Janos Gyarmati-szabo - Co-inventor

Janos Gyarmati-szabo


Lukas Cironis - Principle data scientist

Lukas Cironis

Principal Data Scientist

Leonid Bogachev - Co-inventor and academic lead

Leonid Bogachev

Co-Inventor and Academic Lead

Arshad Mairaj - Director

Arshad Mairaj



Black Swans: What are they and do they really exist?

Black swans are thought to be extremely rare events, and on most occasions, with highly negative consequences. These are usually difficult to predict beforehand but can be foreseen even if there is no historical data detailing similar situations.

While black swans occur in various industries and situations, some examples include health-related matters such as the COVID-19 pandemic, environmental issues – droughts, flash floods, and politically-related issues. As seen with recent World events, unforeseen and unpredictable events can overthrow any preparation and risk-aversion tactics in no time.

Taking all of the above into account, we have to conclude that black swans, unfortunately, do exist.

Now with these thoughts in mind, this is where a reliable and advanced software platform is needed – a solution that includes risk modelling, extreme value modelling, simulation modelling, and synthetic data generation. All these methods have their own functions but are tied together by a program or platform that can assist in better deciphering these usually unpredictable events.

Quantifying uncertainty and managing risk has always been a mainstay of the financial world but the financial crisis of 2008 catapulted risk management software to the fore. The pressure is on for risk management software to keep financial service providers at least one step ahead of catastrophe. But there is a problem – traditional risk models are built with a view on modelling and predicting “typical” scenarios, relying on extrapolation of previously observed patterns and trends. Such approaches struggle to […]

Extreme Value Modelling: Is it necessary in Risk Management?

If we run a model against a hundred years of data, looking to predict the peak for the hundred and first year, the classical model takes the 100 peaks and applies a mathematical rule on how you model these peaks to predict the 101st peak.

The big drawback is the amount of data you waste and the fact that the model needs to be rebuilt if the threshold parameters are changed. This makes the process expensive and a drain on your resources. A better approach developed by 4-Xtra is one where the data isn’t wasted. You draw a high threshold, for example, 95% of the data; so, 5% of the data is very high and 95% is moderate (‘normal’). The moderate values are often called ‘bulk’ and the extreme values are peaks-over-threshold.

The advantage here is that we keep all the values in play, not just one, and model all these peaks.

Until now, the only mathematical approaches available for this have been limited to stationary data. The model created by 4-Xtra answers the question of how to modify this approach in a nonstationary manner to retain threshold stability. When creating statistical models there are three important parameters of the extreme value distribution: shape, scale, and location.

Shape is the most important parameter as it determines one of the three different classes of the distribution. These parameters are constant in the classical model where everything is stationary. Yet in the nonstationary case, […]

Synthetic Data Generation: A Look at How this Benefits Risk Management in Financial Services

Data per se has grown to be an asset in varying industries, never more so than in the financial services sector.

When it comes to making informed business decisions, forecasting, revenue management, and risk management, this resource is vital, it’s not the organisation with the most data that will win, it’s the one that knows how to interrupt it, and how to utilize it that will succeed. Over the years, we have seen its evolution and in more recent times, it has come in the form of synthetic data. This type of data is known as annotated information that is generated by simulations that can take the place of actual data, and it has become an alternative for those that would like to advance their studies for a lesser cost whilst keeping the accuracy of results. While this new breed of data has been around for a couple of decades, its more refined version has become an even more integral part for many financial institutions over the past 24-months.

Although synthetic data is artificial and generated to replicate readily available real-world data, some research studies have shown that the results gathered from using synthetic data for artificial intelligence (AI) model training provide enhanced results compared to actual data from events, objects, or people. It has also been noted that synthetic data is necessary for further development of learning and other potential use cases in the area of risk management or […]

Financial Risk Management: The Rise of Risk Modelling Techniques through Artificial Intelligence and Machine Learning

Over recent years, Artificial Intelligence (AI) and Machine Learning (ML) have made waves in risk management – especially in the financial sector.

From this sector, the banking industry has been at the forefront of this shift from traditional to technological financial risk management, given its dependence on data management and risk assessment, it seems that there is no slowing down. Both artificial intelligence and machine learning have contributed to streamlining and improving customer service through automated chats, fraud detection, optimised forecasting, and other automations.

The ability to forecast, quantify uncertainty and manage risk has always been a staple in the financial world and even more so since the financial crisis of 2008. This event is seen as the reason that development within financial risk management has been catapulted up the agenda as risk departments and financial risk management software providers look to provide the foresight necessary to be one step ahead of any pending disaster.

While the traditional manner of risk management has pioneered this practice, recent times have called for increased automation and information leverage to provide an enhanced level of accuracy and efficiency. Traditional risk models are limited to typical scenarios that rely on the extrapolation of previously observed patterns and trends. This is dangerous because it offers a highly limited scope, and given the dynamic changes in the industry, a better solution must be provided. Aside from this, the traditional approach struggles when it comes to shifting gears, […]

Go to Top