Data-driven decision support for enterprise applications and revenue management (RM) has become highly relevant in recent years as firms face the challenge of integrating data-driven automation in their processes. Specifically, our research group investigates how decision problems can be solved using quantitative methods of operations research and data science in order to improve automated decision-making in the areas of RM and business analytics.

Group Leader: Dr. Rainer Schlosser

Our scope involves (I) the identification of causal relations of underlying dynamics, (II) specific revenue management problems, as well as (III) resource allocation and database tuning problems in close collaboration with the Hyrise database group.

The knowledge of the causal structures within complex systems is crucial to many domains. For example in manufacturing, where causal structural knowledge constitutes the basis of error avoidance in a production process. While domain experts within the company have enough expertise to identify the most common relationships, they will require support in the context of both, an increasing amount of observational data and the complexity of large systems. This gap can be closed by algorithms of causal structure learning (CSL) that derive the underlying causal structures from observational data, e.g., error messages and inline measurements of a production process.

We address open challenges in the context of CSL by improvements in both the application of statistical and probabilistic concepts, and GPU-based acceleration to support a utilization in a real-world context. Read more.

Research Directions & Projects:

• MPCSL - A Modular Pipeline for Causal Structure Learning
• Causal Structure Learning from Heterogeneous Data
• Parallel Execution Strategies for Causal Structure Learning on GPUs
• Application Scenarios with Cooperation Partners
  • Mechanical Engineering
  • Automotive Production

Contact: Christopher Hagedorn, Johannes Huegle, Dr. Rainer Schlosser

Over the years, RM applications have become increasingly difficult to administrate. The number of decisions to control business processes (e.g., dynamic price optimization, inventory management, resource allocation, etc.) have become too complex to be managed manually. Firms are forced to integrate effective automated decision support systems in order to be still profitable. However, while automated decision-making has enormous potential, it is also challenging to derive optimized decisions as most RM applications are typically highly complex stochastic dynamic optimization problems. Nevertheless, the overall vision is a self-driving decision support system, which automatically analyzes sales data and optimizes decisions in an explainable manner.

In general, decision problems can be described by given performance criteria, admissible decisions, constraints, and data-driven estimations of the interplay of decisions on performance. Further, every application has its own specifics, which can be exploited to solve a problem effectively by using suitable optimization techniques. In particular, we are interested in finding robust solutions for uncertain and changing environments (read more).

Research Directions & Projects:

• Risk-averse Control of Markov Decision Processes
• Approximate Dynamic Programming & Reinforcement Learning
• Circular Economy & Sustainable Revenue Management
• Dynamic Pricing on Amazon Marketplace
• Price Wars: An Interactive Simulation Platform

Contact: Dr. Rainer Schlosser, Alexander Kastius, Martin Boissier

Research Directions & Projects:

• Robust Database Optimization with Stochastic Workloads
• Memory-efficient Fragment Allocation & Load Balancing
• Dynamic Risk-aware Index Selection
• Spatio-temporal Data Management
• Self-tuning Databases

Contact: Martin Boissier, Stefan Halfpap, Jan Kossmann, Keven Richly, Dr. Rainer Schlosser

We offer lectures, seminars, and projects on data-driven decision-making in enterprise applications. For HPI master students we also provide a varity of master theses topics. 

On a more applied level, we also offer a one-year bachelor project in cooperation with the industrial partner SAP SE in the winter semester. This year the project is about simulating markets with self-learning agents in the ReCommerce industry (cf. sustainability & circular economy).

• Schlosser, R.: Keynote presentation at ICORES 2023, Feb 2023, Lisbon, Portugal
• Kastius, A. (with N. Kiele, R. Schlosser): "Multi-Agent Dynamic Pricing Using RL and Asymmetric Information", International Conference on Operations Research (OR 2022), Sep 2022, KIT, Germany
• Schlosser, R. (with A. Kastius): "A Conceptual Framework for Studying Self-Learning Agents in Recommerce Markets", International Conference on Operations Research (OR 2022), Sep 2022, KIT, Germany

• Schlosser, R. (with A. Kastius): "Stochastic Dynamic Pricing under Duopoly Competition with Mutual Strategy Adjustments", International Conference on Operations Research (OR 2021), Sep 2021, Bern, Switzerland
• Kastius, A. (with R. Schlosser): "Towards Transfer Learning for Revenue and Pricing Management", International Conference on Operations Research (OR 2021), Sep 2021, Bern, Switzerland
• Huegle, J.: "Causal structure learning for heterogeneous data characteristics of real-world scenarios", European Causal Inference Meeting (EuroCIM 2021), Sep 2021, Lightning Talk, Virtual
• Huegle, J. (with C. Hagedorn, M. Perscheid, H. Plattner): "MPCSL - A Modular Pipeline for Causal Structure Learning", 27^th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (KDD 2021), Aug 2021, Virtual
• Schlosser, R.: "Scalable Relaxation Techniques to Solve Stochastic Dynamic Multi-Product Pricing Problems with Substitution Effects", European Conference on Operational Research (EURO 2021), July 2021, Athens, Greece
• Kastius, A. (with R. Schlosser): "Dynamic Pricing under Competition using Reinforcement Learning", European Conference on Operational Research (EURO 2021), July 2021, Athens, Greece
• Schlosser, R. (with J. Gönsch): "Risk-Averse Revenue Management using Mean-Variance and Mean-Semivariance Optimization", INFORMS Revenue Management & Pricing Conference, June 2021, Johns Hopkins Carey Business School, USA
• Hagedorn, C. (with J. Huegle): "GPU-Accelerated Constraint-Based Causal Structure Learning for Discrete Data", SIAM International Conference on Data Mining (SDM 2021), Apr 2021, Virtual
• Schlosser, R (with Y. Kaminsky, T. Maltenberger, M. Pörschke, J. Westphal): "Pricing Competition in a Duopoly with Self-adapting Strategies", 17^th International Conference on Operations Research & Enterprise Systems (ICORES 2021), Feb 2021, Vienna, Austria

Our research with over 60 publications has been published in renowned OR Journals (EJOR, JEDC, COR, IJPE, IJPR, DGAA, JRPM, JIMS), distinguished data science conferences (KDD, IJCAI, RECSYS, SDM), and leading computer science venues (VLDB, ICDE, EDBT, DAPD, CIKM, SSDBM).