Foto: Simple ML
Machine learning workflows made easy
Anyone who wants to use the latest machine learning (ML) techniques faces a major challenge. Even supposedly simple tasks like traffic forecasting raise a number of questions: Where can I find suitable vehicle data for my forecast? How can I use them in combination with weather and map data? Which ML procedure should I choose? How do I describe my ML workflow? How good is my prognosis? This can quickly frustrate especially newcomers to data processing and machine learning, even if they have sufficient expertise in their own field.
In the project Simple-ML the L3S is working together with the Rheinische Friedrich-Wilhelms-University Bonn, the AIM Agile IT Management GmbH and the PROJEKTIONISTEN GmbH to make ML processes more easily accessible for a broad user group. This includes the provision and description of suitable data sets, help and tips for the selection of ML models as well as the explanation and interactive visualization of the results – all based on semantic data profiles, robust workflow specifications with a domain-specific language and scalable implementations. A special focus is on the implementation of ML procedures in the domains of mobility and logistics, as open heterogeneous data are increasingly available in these areas in particular. They require specialized operations and visualizations that go beyond the use of traditional ML tools.
Simple-ML should therefore facilitate access to machine learning, make ML more comprehensible and offer interaction possibilities, so that, for example, all subtasks on the way to a traffic forecast are successful without significant ML expertise and a robust, comprehensible and efficient ML workflow is created.
Elena Demidova is project coordinator of Simple-ML and researches at L3S in the areas of AI, data analytics and mobility.
Simon Gottschalk is a PhD student at L3S, researching knowledge graphs, events and semantic data profiles.