Tag Archives: data science

Research with Impact – Department of Computer Science and Information Systems

The impact case studies submitted by our Department of Computer Science and Information Systems were rated 67% 4* (world-leading) and 33% 3* (internationally excellent), demonstrating the significant effect that Birkbeck research is having on populations both in London and around the world. More details about each case study are given below; they can be found in full on the REF website.

Enhancing Vehicle Deployment Strategies at the London Ambulance Service

When somebody is critically ill, the speed at which an ambulance can reach them can be a matter of life or death. Ensuring ambulance coverage across a given region is therefore critically important, and an important part of maintaining this coverage is understanding how long it will take an ambulance to get from point A to point B. This is particularly true in London, where the London Ambulance Service (LAS) works across the UK’s most densely populated area, requiring ambulances to contend with high volumes of traffic as they carry out their work.

Working with PhD student Marcus Poulton, who was then employed by the LAS, Birkbeck researchers George Roussos and David Weston (and their collaborator, Anastasios Noulas of the NYU Data Science Institute) were able to improve the mapping systems used to predict ambulance travel times around the capital. Their software, tailored for the specific travel patterns of emergency vehicles (which are, for example, able to use some road sections that ordinary vehicles cannot) are 80% more accurate than the previous best-in-class and are integral to the Geotracker software used by LAS despatchers to plan the movement of the service’s vehicles around the capital. According to LAS staff, ‘The work from Birkbeck has changed the way we map and deploy ambulances for the better, and that means helping to save lives.’

Virtual Knowledge Graphs in industry and the public sector

Virtual knowledge graph technology is used in situations where an organisation holds a high volume of complex data across databases and repositories that may not have been designed to work together. By defining and mapping these disparate pieces of information, a VKG allows ordinary users to search across these multiple datasets in a natural, straightforward way. The benefits to this technology can be enormous, as staff members are able to find in minutes information that previously required the involvement of IT specialists over several days or weeks.

Birkbeck researchers Michael Zakharyaschev and Roman Kontchakov are at the forefront of VKG development. Their work is cited in the World Wide Web Committee’s definition of OWL 2 QL, a subset of the Web Ontology Language (OWL) that is used for writing virtual knowledge graphs and which is in use within such systems worldwide. Zakharyaschev and Kontchakov have also contributed directly to the development of one specific VKG system, Ontop, based at the University of Bozen-Bolzano in Italy. Organisations around the world are using Ontop to carry out complex data analysis in a huge variety of fields, from the Norwegian oil and gas industry to Brazilian cancer research, bringing economic, social, political and health benefits to the populations that they serve.

Global Standards for Smart City infrastructure: Entity Identification Systems

Smart cities, which use information and communications technology to run key services like transport and sanitation, are growing in number around the globe. To function effectively, smart cities need a robust system of entity identification that allows them to distinguish unique items such as artefacts, products, and buildings. Worldwide, these systems are numerous, often incompatible, and frequently in direct competition with one another, which makes it difficult to transfer learning from one city to another and therefore slows down innovation. This matters because smart cities have been shown to be more efficient, healthier, and more environmentally sustainable than their traditional counterparts, with a 2018 report from McKinsey stating that ‘smart technologies can reduce fatalities by 8-10 percent, accelerate emergency response times by 20-35 percent, shave the average commute time by 15-20 percent, reduce the disease burden by 8-15 percent, lower greenhouse gas (GHG) emissions by 10-15 percent, and reduce water consumption by 20-30 percent’.

As a key member of the International Telecommunications Union’s study group on smart cities (SG20), George Roussos worked to develop a worldwide standard for global information infrastructure: ITU-T Y.4805. This specifies the functionality for federated entity identification services in Smart City applications, ensuring that such systems are interoperable and secure. As part of a package of ITU standards around the development of smart cities and the internet of things, Roussos’s work has underpinned smart city implementations around the world, notably in China and Africa, and feeds into the Smart Sustainable City standards towards which cities from Montevideo to Dubai are working.

Further Information



Forecasting the Trajectory of an Epidemic

Mark Levene is Professor of Computer Science in Birkbeck’s Department of Computer Science and Information Systems. He shares insights from research into modelling the waves of an epidemic.

Epidemics such as COVID-19 come in “waves”, although the precise definition of a wave in this context is somewhat elusive.  A standard way to model the epidemic is as a time series that records, say the number of daily hospitalisation or deaths, and these can be plotted to view the progress of the epidemic.

Waves in the time series span from one valley to another with a peak in between them. The shape of an individual wave can be modelled as a statistical distribution and several waves can be sequentially combined. More often than not waves are not symmetric, that is, the rate at which, say hospitalisations, increase is not the same rate at which they decrease once the peak of the wave has been reached. This non-symmetrical nature of a wave is called its skewness.

To take into account the skewness of epidemic waves we introduce the skew logistic distribution, which is a novel yet simple extension of the symmetric logistic distribution widely used in the modelling of epidemic data.

To validate our model, we provide a full analysis of the first four waves of COVID-19 deaths in the UK from the 30 January 2020 to 30 July 2021.

Our results show a good fit to the proposed skew logistic distribution, and thus could potentially augment existing more established models that are being used to forecast the trajectory of an epidemic.

Our findings have been published in MDPI Entropy.

Further Information