Data Science Research Group: Home

DataScienceGroup @ DSV

Welcome!

The Data Science Research Group at the Department of Computer and Systems Sciences (DSV) of Stockholm University focuses on core data science resarch as well as on application areas where data science can assist in providing useful insights for decision making. The group focuses on the formulation of novel data science problems and the development of algorithmic methods and methodological workflows for achieving scalable solutions in core application areas within social sciences and humanities.

News

February 21, 2021
[Data Science @ DSV - PhD Workshop: Mar 26 2021] The Data Science group is organizing a PhD Workshop where our PhD students will present their ongoing research. More information is available here.
December 12, 2020
[Postdoc in Data science - Deadline [extended]: Jan 25 2021] The Data Science group is looking for a postdoctoral researcher in data science. More information is available here.
December 06, 2020
[AI Masters Program - Deadline: Jan 15 2021] Our one-year Master’s Program in AI is accepting applications. More information is available here.
December 05, 2020
[Research Assistant in Data Science - Deadline: Dec 28 2020] The Data Science group is looking for a research assistant in data science. More information is available here.

Older posts…

Collaborators

Spotify

SPV (Swedish Pensions Agency)

Research Profile

The group has three main focus research themes: sequential and temporal mining, explainable and federated learning, and machine learning applications. For a comprehensive list of our research topics please visit our research page.

Sequential & Temporal Mining

We focus on developing methods for searching and mining rich and complex data sources, with emphasis on sequential and temporal data, as well as text. In particular, we are interested in defining temporal abstractions and extracting high-utility features for clustering or classification of sequential and temporal data sources, such as univariate and multivariate time series, event sequences, and text.

We give particular emphasis on methods and workflows for explainable machine learning. We explore both model agnostic and model specific solutions, as well as counterfactual explanation formulations for sequential data variables, images, and text. Our main goal is to provide scalable and distributed solutions for maintaining good trade-offs between predictive performance and explainability. Moreover, we are interested in solutions that can function in a distributed manner without the need for data exchange.

Explainable & Federated Learning

Machine Learning Applications

Our methods and solutions are motivated by real-world applications and use-cases. The group has particular expertise on mining and model understanding from healthcare and medical data sources. In addition, we have established strong expertise on predictive maintanance and integrated vehicle management. Finally, we are interested in financial data, enironmental data, as well as data emerging from immersve technologies, such as virtual reality.