Natural language processing
leading company in the field of microcredit
WebTrekk and Uebermetrics
Professor Lessmann's research focuses on machine learning and artificial intelligence (MLAI) methodologies and their use cases in managerial decision support. Areas of interest include but are not limited to explainable AI, causal/interpretable machine learning, natural language processing, predictive analytics and time series forecasting.
He is specialized in MLAI applications in the broad scope of marketing and risk analytics. He is also actively involved in knowledge transfer, professional education and consulting projects with industry partners ranging from start-ups to global players and non-profit organizations.
- Prototyping, evaluation, and benchmarking of ML/AI-based prediction and decision models
- Development of PD, LGD, and EAD forecasting models
- Processing of textual data for knowledge extraction
- Assessment of the information value of commercially available data sources (e.g., credit bureau data) for predictive modeling
- Estimation of personalized, individual-level treatment effects
- Professional education in ML/AI, data science and business analytics
FinTech - Based on concepts of Bayesian statistics and semi-supervised machine learning, Professor Lessmann and his team have developed a system to solve the problem of inference rejection in the lending industry. Their solutions provide significantly better risk scorecards and make it easier for lenders to predict the predictive power of a scorecard (e.g., the PD) in operation. The project was conducted with a leader in the microcredit space.
E-commerce - Together with WebTrekk and Uebermetrics, Lessmanns working group is developing an AI-driven web controlling system that examines web metric time series (page views, visits, bounce rates, checkouts, sales, etc.) and identifies the causes of observed patterns (e.g., anomalies). To this end, their system integrates deep causal detection algorithms based on reinforced learning and xAI methods. To extend the scope of their root cause explanations, they also monitor various social media streams and use advanced NLP algorithms for event detection. The unique combination of these AI concepts provides web controllers and store managers with advanced insights into their business operations.
Digital marketing - Lessman and his team have developed advanced causal machine learning models for real-time targeting of e-coupons and other digital marketing stimuli. Their optimal (i.e., profit-maximizing) methodology includes both single and multiple treatment scenarios and leverages traditional supervised learning algorithms to enable cost-effective deployment.
Professor Lessmann has taught a number of professional education courses in subject areas such as credit risk modeling, fraud detection, advanced analytics, and machine learning, to name a few. Some of these courses are offered as part of the SAS Business Knowledge Series, but are also available on the R and Python platforms. Requests for professional development in the areas of his expertise are welcome.
- funding by various bodies including DFG, DAAD, IBB (e.g. Transfer Bonus, ProFIT)
Prof. Dr. Alan Akbik conducts research in the field of machine learning (ML) and natural language processing (NLP). His goal is to enable machines to capture, understand, and use natural language like a human.
To this end, he has developed one of the world's leading deep learning frameworks for NLP, which is already being used in over 1,000 research and industrial projects.
Consulting in:
- Deep Learning
- Natural Language Processing (NLP)
- Machine Learning (ML)
- 3 Patents at IBM Research
- Zalando Outstanding Achievement Award
workshop on the analysis of Twitter data for Stiftung Wissenschaft und Politik (SWP)
Professor Jäschke and his team develop and optimize methods in the areas of Big Data and Machine Learning, especially in the aplication fields of Natural Language Processing, Social Bookmarking and Recommendation Systems. This includes the Collection (e.g. through focussed Crawling), Compilation, Annotation (e.g. by means of Crowdsourcing) and Curation of suitable records (data sets).
Further, it includes the Adaptation and Improvement of appropriate Algorythms (e.g. Named Entity Recognition, Classification, Clustering, Information Extraction, etc.) culminating in the development of web-based analysis platforms.
- Hadoop-Cluster
- GPU calculator
- diverse datasets