Most mathematical approaches used in risk assessment and critical decision making rely too heavily on 'big data' and statistical analysis. Graphical models, known as Bayesian Networks, could help address this problem, but until now only highly trained experts could build and use these models.
Professors Norman Fenton and Martin Neil from Queen Mary University of London have created AgenaRisk, a new piece of software that makes Bayesian Networks more accessible.
The software allows users to combine data with expert knowledge to improve the decision-making process, and has gained over 4,500 new users since 2014.
What are Bayesian Networks?
Most mathematical approaches to risk assessment and decision analysis rely on 'big data' – huge, pre-gathered data sets that often include hundreds of thousands of information points, which is often analysed using computers to reveal patterns, trends and associations. However, this means they are not always appropriate, especially in uncertain situations where the data may be imperfect, or even unknown, as they rely on assumptions based on statistical analysis. On occasions where rare or new risks may occur, there is often little or no relevant data to use. Even when looking at common risks with extensive historic data, crucial data on contributing causes and explanatory factors may often be missing.
Bayesian Networks (BNs) can address this problem. A BN is a graphical model that represents knowledge about an uncertainty. The models contain nodes, which correspond to a random variable, and edges, which represent a conditional probability for each random variable.
For example, a Bayesian network structure for a medical diagnosis can look at a patient’s symptoms and relevant background information to determine what type of illness they may be suffering from. If a certain disease is found only in one country, which the patient hasn’t travelled to, then the probability the patient has that particular illness would be very low.
While BNs can help address problems with existing approaches to risk assessment and decision making, they also have their own issues.
As they often depend on 'machine learning' approaches, they work by finding patterns in existing data and produce little new insight. Therefore, the machine learning approach is only suitable when large volumes of data are available and most of the relevant relationships are already known.
In addition to this, without a statistical background and expert knowledge of BNs, it can often be very difficult for people to develop and use the models to solve real world risk problems.
Creating software based on Bayesian Networks
To address both issues, Professors Fenton and Neil drew on their extensive research with BNs to develop a range of techniques to make it easier to use the models for risk management and decision making.
By combing data with expert knowledge, they developed an alternative 'smart data' approach to produce a new type of BN model. This research was then incorporated into a new software system, AgenaRisk. In 1998 Professors Fenton and Neil founded Agnea Ltd, and created Agena Risk from this company. The software combines data and expert knowledge to improve decision making and risk assessment predictions. Users can use the software for safety and reliability concerns, solving cost benefit problems and satisfying regulators.
AgenaRisk is currently being used in a variety of fields, from helping to model and predict cybersecurity risk, to supporting medical decision making for trauma patients.
Supporting clinical decision making
Professors Fenton and Neil worked with trauma surgeons at the Royal London Hospital, one of the busiest trauma centres in Europe, to develop two clinical decision support (CDS) tools to construct 'causal networks'. These would reproduce the reasoning pathways that clinicians use to determine if a patient is likely to be at risk of a particular condition or outcome.
The first tool has been shown to be as accurate as specialised blood tests for coagulation. Unlike blood tests, it doesn't require the use of sophisticated laboratory equipment. This allows the clinical team to offer damage control resuscitation to patients who need it, and spare patients who don't need it the potential side effects of this treatment.
The second tool, named the Limb Salvage CDS tool was designed to help doctors make decisions about how to save the limbs of injured patients.
These decision-support tools, designed to meet the needs of time-poor clinical decision makers working under pressure, will improve patient care – especially so in under-resourced settings where sophisticated laboratory kit or the absence of senior surgical support threatens optimal patient care.— Nigel Tai, Consultant Trauma and Vascular Surgeon at the Royal London Hospital
Modelling and predicting cybersecurity risk
AgenaRisk is also in use by five US and Indian technology companies to model and predict cybersecurity risk. BARD, the international collaborative project funded by the US Intelligence and Research Agency, used AgenaRisk across all its sites to tackle a variety of intelligence related problems. The companies looked at problems where their agents had a variety of evidence from multiple sources all with varying accuracy, some of which could be contradictory. Fenton and Neil worked as consultants to develop intelligence related models, and trained users to build and run these models.
Since the end of the project Monash University in Australia has been using the AgenaRisk application programming interface (API) as a platform for its own web-based intelligence modelling system.
Schools, institutes and research centres
School of Electronic Engineering and Computer Science
With a 130-year history, our School offers a vibrant, multi-disciplinary learning and research environment. Our enthusiasm for research defines our programmes, keeping our teaching exciting and relevant.