TimeXAI

The research field of Explainable Artificial Intelligence (XAI) aims to make Artificial Intelligence (AI) more transparent and interpretable. Especially in safety-critical applications, it is essential to be able to understand the algorithmic decisions of black-box models. XAI methods can generate explanations for the decision-making processes of AI models that are understandable to humans. On the one hand, these explanations provide valuable insights for the development and optimization of AI models, and on the other hand, they help build trust among users of such systems. The goal of the junior research group “TimeXAI” is to develop a novel XAI method for time series classification, with a strong focus on the human interpretability of the explanations. Using various evaluation approaches, the strengths of existing XAI methods are identified and combined into a new method. In recent years, a wide range of promising XAI methods has been developed in the field of computer vision. However, XAI approaches for AI models in time series classification are underrepresented, even though the demand for such solutions is high and will continue to increase significantly due to the rapid growth of sensor data. Therefore, the specialized methods developed within “TimeXAI” offer great potential for numerous applications and significant innovation advances. The applicability and interpretability of XAI methods are investigated in two initial application domains:

Cardiology

The application domain of cardiology is a collaboration with, among others, Prof. Dr. Michael Guckert (THM), Prof. Dr. Samuel Sossalla (Kerckhoff Clinic Bad Nauheim), Prof. Dr. Om Prakash Vyas (Indian Institute of Information Technology Allahabad, India), and CRS medical GmbH (Aßlar).

Smart Home

The Smart Home application domain is a collaboration with the startup Veli GmbH. Veli is a spin-off of the University of Kassel, and its product enables elderly people to live independently in their own homes for as long as possible by analyzing water and electricity consumption using AI-based sensor data.

The junior research group TimeXAI is funded for a duration of three years by the Federal Ministry of Research, Technology and Space (BMFTR) under the funding program ExperTeam4KI. The project receives a total funding of €1.09 million (funding reference: 01IS24063).