You are here: Home Research Machine Learning in Business …

Machine Learning in Business Analytics

The research group "Machine Learning in Business Analytics" performs multi-disciplinary research at the intersection of computer science, economics, and data science. The activity of our group is directed towards a broad selection of topics including prediction and optimization problems from financial markets and operations research.


Leadership

Bernhard Lutz

Bernhard Lutz, Dr.
Coordinator

Members of the Research Group

Janin Kortum

Janin Kortum

Sergej_Levich_.jpg

Sergej Levich

Antal Ratku

Antal Ratku

Tano Müller

Tano Müller

Jannik Schäfer

Jannik Schäfer

Torsten Mörstedt

Torsten Mörstedt

hans_christian_schmitz_res.jpg

Hans Christian Schmitz

 

 

 

 



Research Focus:
Reinforcement Learning in Operations Research

We consider optimization problems from the field of operations research. Among others, we consider scheduling problems in the automotive industry and investment decisions based on noisy predictions about future returns and risks. For this purpose, we develop novel methods based on reinforcement learning to learn optimization via "trial-and-error". In addition, we address current challenges of training a reinforcement policy directly in the real-world.


Research Focus
: Financial Text Mining

The availability of information forms the basis for financial decision-making. We use modern algorithms like word embeddings and multi-instance learning to get a deeper understanding of how corporate disclosures are perceived by investors. In addition, we develop trading strategies based on signals extracted from financial documents.


Research Focus
: Covariance Estimation

The minimum variance portfolio is a strongly desired portfolio allocation. The weights of the GMV portfolio by Markowitz (1952) only depend on an accurate estimate of the future covariance matrix. However, the sample covariance generally provides a poor estimate of future investment opportunities. We employ tools and methods from machine learning for non-linear shrinkage of the sample covariance matrix to obtain more robust estimates.