Tagged / Alan Turing Institute

Insights from the Alan Turing Institute Data Study Group

Stepping into the world of data science at the Alan Turing Institute (ATI) Data Study Group from September 9-13, 2024 was an exhilarating experience. As the national institute for data science, the ATI’s strong connections to academia and industry set the stage for a week of intensive collaboration and problem-solving.

Joining the Transport for London (TfL) project within the hackathon-style event, I found myself amidst an eclectic team of 11 individuals, each bringing a unique set of skills and backgrounds to the table. The project’s challenge of identifying physical assets on the London Underground from point cloud data presented a thrilling opportunity to apply our collective expertise to a real-world problem.

The task at hand was no small feat – analysing point cloud data to pinpoint key track features along the London Underground network. With collected image data, our goal was to automate the detection and classification of critical track components such as sleepers, rails, signalling equipment, and more. Our team’s approach was multifaceted, involving rigorous preprocessing, segmentation model training, and advanced data analysis techniques. By leveraging tools like U-Net and SAM 2 for image segmentation and employing post-processing methods to extract valuable insights from the predicted masks, we made strides towards achieving our objectives.

As an individual who recently completed a PhD thesis on “Complex Urban Road Networks: Static Structures and Dynamic Processes,” this opportunity to apply my research expertise in a practical setting was both challenging and rewarding. The seamless blend of academic knowledge and hands-on problem-solving during the ATI Data Study Group not only expanded my technical skills but also reinforced the importance of interdisciplinary collaboration in tackling complex data challenges.

Thanks to The Alan Turing Institute,  especially TuringDSG organisers for an incredible opportunity. I would like to extend my gratitude to my PhD supervisor Dr. Wei Koong Chai for supporting my professional and personal development.

Assemgul Kozhabek (Computing Department, SciTech  )

Bournemouth Academic invited to present at Developing Social Data Science Methodologies workshop within the Alan Turing Institute

Katarzyna Musial-Gabrys was invited to present her work on complex social networks during the upcoming workshop organised by the Alan Turing Institute within the Foundation of Social Data Science initiative.tag_claud_2011

The Alan Turing Institute was established in 2015 as the UK national institute for the data sciences in response to a letter from the Council for Science and Technology (CST) to the UK Prime Minister (7 June 2013), describing the “Age of Algorithms”. The letter presents a case that “The Government, working with the universities and industry, should create a National Centre to promote advanced research and translational work in algorithms and the application of data science.” (https://www.gov.uk/government/publications/the-age-of-algorithms).
Katarzyna’s presentation will contribute to shaping the portfolio of research challenges to be addressed within the Alan Turing Institute.

Title of Katarzyna’s talk: Methodological challenges in data aggregation in complex social networks.

Abstract of the talk:
For the first time in history, we have the possibility to process ‘big data’ (gathered in computer systems) about the interactions and activities of millions of individuals. It represents an increasingly important yet underutilized resource because due to the scale, complexity and dynamics, social networks extracted from this data are extremely difficult to analyse. There is no coherent and comprehensive methodological approach to analyse such networks which is crucial to advance our understanding of continuously changing people’s behaviour.
One of the methodological challenges is to cope with the variety of available big social data. This data comes from multiple systems (email, instant messengers, blogs, social networking sites, google searches, YouTube, etc.); in each system user can have one or more accounts; this data describes different types of activities (commenting, sharing, messaging, calling, etc.) and relationships (direct, quasi-direct and indirect). In order to be able to effectively process gathered data using data science approaches we need to develop new methodology that will focus on the multirelational (more than one type of connections in a network) character of data.
In general, there are two methods to do that: (i) analyse each relation type separately and then combine results from different layers or (ii) merge all relation types in one layer and analyse this newly created layer. Both approaches require effort in terms of redefining existing network analysis techniques. Analysing each network separately means that methods for combining results from different layers need to be developed. Merging some/all connection types into one heterogeneous relation means that a new approach for aggregation of data from different layers is required. Only by developing rigid approaches to data aggregation, the analytics task can be performed.

If you are interested and you would like to get some further information please contact kmusialgabrys@bournemouth.ac.uk.