Meeting: 15th of March 2024

Venue: Goldsmiths, University of London, New Cross, London SE14 6NW. 

Room: RHB 137a in Richard Hoggart Building

17:00 to 17:45    Annual General Meeting (AGM) 

18:00 to 19:00    Seminar of British Data Science Society (BDSS)  (Open event)

Title: Predicting psychosis in high-risk patients: An overview of current work at the Institute of Psychiatry, Psychology & Neuroscience (IoPPN)

Speaker: Professor Daniel Stahl, King’s College London 

Clinical prediction models are the backbone of modern precision medicine. This presentation will offer an overview of the importance of clinical prediction modelling and the potential impact these models can have on patient care and decision-making processes. I will present some current research on predicting the risk of psychosis in people with severe mental health problems using electronic health records. In addition to introducing the models, I will also discuss the challenges of implementation and avenues for further research.

This work was supported by the NIHR Maudsley Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King’s College London. The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care.

Bio: Dr Daniel Stahl is a Professor of Medical Statistics and Statistical Learning and lead of the NIHR Maudsley BRC "Prediction Modelling" group at King’s College London. His interest is applying statistical and machine learning methods to develop and implement robust risk prediction and treatment outcome models. He is also interested to identify predictors, mediators, and moderators of treatment success and applying model-based cluster analysis methods to identify subgroups among psychiatric patients. Prof Stahl currently holds the role of Deputy Head of Department of Biostatistics and Health Informatics at KCL, and Deputy Lead of the NIHR Maudsley BRC theme “Trials, Genomics and Predictions”.

Meeting: 6th of October 2023


Our meeting on the 6th of October will happen at the University of Essex, room  1N1.4.1. You can find a map here.

13:00 Anomalous clustering at various data formats (on-site)
Boris Mirkin (NRU HSE Moscow RF, University London UK)

Anomalous clustering is a method for extracting clusters one-by-one. It is an extension of the Principal Component Analysis method to zero-one matrix factorization settings. After a brief overview of various versions of the method, including its extensions to similarity data, spatial data, and fuzzy clustering, I am going to concentrate on a most recent development, a triple-stage application of the approach to the analysis of spatial-temporal patterns in a coastal oceanic phenomenon of upwelling (see Nascimento et al. 2023).


Nascimento, S., Martins, A., Relvas, P., Luís, J. F., & Mirkin, B. (2023). Core-shell clustering approach for detection and analysis of coastal upwelling. Computers & Geosciences, v. 179, 105421.

14:30 Temporary GPs and the effects on patients' health outcomes (online)
Cristina Orso (University of Insubria)

The impact of temporary work has been studied extensively in the literature, but little is known about the implications of temporary work in the healthcare sector. In this scientific paper, we investigate the impact of locum doctors on patients’ satisfaction and prescription behaviours using a unique dataset that matches the information on temporary contracts for 6781 healthcare practices in England from 2017 to 2022, along with patient satisfaction ratings and psychotropic medication drug prescriptions.

We employ panel data techniques that leverage both the cross-sectional and temporal dimensions of the dataset to analyse the relationship between locum doctors and mental health outcomes. Our findings indicate that patient satisfaction is lower in practices with more temporary job contracts. This result supports our hypothesis that patients may prefer a less precarious relationship with their healthcare providers.

We also find a positive association between the higher share of locums and psychotropic medication prescriptions, while there is a negative association with antibiotic and infection prescriptions. The reduced time that locums may have to engage with their patients may incentivize them to prescribe, or possibly over-prescribe, psychotropic medications. This suggests that locum doctors may have an adverse impact on the appropriateness of treatments for patients.

Our results have significant implications for policy interventions aimed at increasing the flexibility of the labour market in the healthcare sector. Such reforms should also consider the economic and social costs of the changes, including the psychological well-being of patients and the appropriateness of their treatments. Our study highlights the importance of ensuring that temporary work arrangements in healthcare do not compromise the quality of patient care and treatment outcomes.

15:00 Pareto optimization in applied data science – formalizing the exploration of optimal trade-offs (online)
Julia Handl (University of Manchester)

Practical data science problems typically require the consideration of a set of criteria, which may reflect stakeholder requirements, prior domain knowledge or constraints associated with other component of the modelling pipeline. In this talk, I will discuss the limitations of traditional (single-objective) data science tools in modelling these problem aspects, and highlight Pareto optimization as a more appropriate modelling framework. Using examples from a range of data science applications, I will highlight the potential of Pareto optimization in formulating and addressing both issues of model generation and / or subsequent model selection.

15:30 The power of AI recommendations: from consumer decisions to industrial safeguards (online)
Sabrine Mallek (ICN Business School)

In today's digital landscape, artificial intelligence (AI) is at the forefront, from changing how we shop online to making industrial factories safer. However, a significant challenge lies in seamlessly integrating AI solutions with traditional methods, often leading to inefficiencies in various sectors. Exploring AI is a journey; how do information systems (IS) and computer science offer keys to the same lock? This talk will delve into the influence of social recommender system such as voice assistants on consumer behaviour, emphasizing factors that drive their impact on products purchases, rooted in IS research. As we transition to a computer science perspective, we will introduce a recommender system approach tailored for industrial risk management, highlighting its real-world effectiveness in the oil industry. Through this journey, attendees will discover the multifaceted applications of AI, while gaining insights into the diverse methodologies for approaching AI research.

Meeting: 8th of October 2022

On some Classification and Survival Machine Learning Approaches to Dementia Risk Prediction, Diagnosis and Biomarker Discovery.
Daniel Stamate

Modelling Classification in Information Systems Engineering.
Sergio De Cesare

Bootstrap evaluation of unsupervised hierarchical statistical learning.
Berthold Lausen

The mediating role of public emotions in managing COVID-19 pandemic.
Meichen Lu and Maged Ali

On the number and meaning of k-means iterations on Gaussian clusters.
Renato Cordeiro de Amorim

David Wishart BCS Poster Awards

Yiyuan Han
1st Prize, £500.

Hao Zhou
3rd Prize (tie), £100.

Haidee Sau Man Tang
3rd Prize (tie), £100.

The agenda for this, with notes on outcomes, can be found here.