Machine learning (ML) has become an important methodological toolkit for empirical research in business and economics. The course provides an introduction to modern ML, emphasizing its use in empirical research. It addresses participants with heterogeneous prior experience.

Machine learning (ML) has become an important methodological toolkit for empirical research in business and economics. It enables researchers to analyze complex, high-dimensional data, develop accurate predictive models, learn representations from data, and address research questions that are difficult to study with conventional statistical methods alone. At the same time, the growing availability of powerful software libraries, foundation models, and generative AI shifts attention from merely implementing ML algorithms toward selecting appropriate methods, designing credible empirical studies, critically evaluating model outputs, and understanding the boundaries of predictive analysis.

The course provides an introduction to modern ML, emphasizing its use in empirical research. It addresses participants with heterogeneous prior experience. To that end, we start from the foundations of supervised learning and model evaluation. Covering these foundations, we progress rapidly toward contemporary developments such as neural representation learning, transformers, foundation models, modern methods for tabular data, and machine learning for time series and longitudinal data.

Throughout the course, methodological concepts are connected to questions of empirical research design. Participants discuss how ML can support prediction, measurement, pattern discovery, and other empirical research tasks; how ML studies should be evaluated; and under which circumstances predictive modeling alone is insufficient. The latter perspective motivates an introduction to related research directions including causal inference, algorithmic decision-making, and responsible machine learning.

After completing the course, participants should be able to:

• understand the main principles underlying contemporary supervised machine learning;
• select and critically evaluate suitable ML approaches for empirical research problems;
• design appropriate validation and model-comparison strategies;
• understand the main ideas behind neural networks, representation learning, transformers, and foundation models;
• recognize methodological challenges arising from tabular and longitudinal data;
• interpret and critically assess predictions from complex ML models;
• identify research questions for which ML can provide substantive or methodological value;
• distinguish predictive questions from causal and decision-oriented research questions and recognize important limitations of predictive ML.

The course targets Ph.D. students who are new to machine learning as well as participants who already have practical ML experience and want to deepen their understanding of recent methodological developments and their implications for empirical research.

Lecturer
Prof. Dr. Stefan Lessmann
Humboldt-Universität zu Berlin
stefan.lessmann@hu-berlin.de

Language

Englisch

Location

aQua-Institut GmbH für angewandte Qualitätsförderung und Forschung im Gesundheitswesen GmbH
Maschmühlenweg 8 – 10, Raum Emmy Noether I + II
37073 Göttingen