VHB - ProDok Machine Learning

Activity: Participation in or organisation ofWorkshop, seminar or course (Participation in/Organisation of)

Description

The course provides a comprehensive overview of the state-of-the-art in machine learning and its applications in business and economics. To that end, the course splits into three parts. Part I revisits fundamental concepts of machine learning including approaches for unsupervised, supervised, and reinforcement learning. Further topics of the first part comprise a discussion of the connections between machine learning and more traditional data analysis paradigms such as statistics and econometrics and the fundamental differences between data-driven models for descriptive, explanatory, predictive, and normative decision support. The objective of Part I is to (re-)introduce selected fundamentals of machine learning and relevant machine learning algorithms. We emphasize techniques for supervised machine learning, which are most relevant for machine learning-oriented research in business and economics. Part II examines recent developments in the scope of deep learning using artificial neural networks. Promising autonomous feature extraction, deep learning advances conventional approaches for machine learning toward artificial intelligence. Deep learning has become the de facto standard for processing large unstructured data sources such as text and images. Following an introduction of deep neural networks, the course concentrates on approaches for processing sequential data using the example of textual data. Considering a piece of text as a sequence of individual tokens (i.e., words) ensures that the techniques covered in the course are readily applicable to other types of sequential data such as time series. At the same time, participants have an opportunity to explore state-of-the-art approaches for natural language processing. Part III covers selected topics in machine learning research. (Deep) machine learning algorithms have proven their ability to process large and heterogeneous high-dimensional data sets. Emphasizing scalability as a design principle, machine learning has to a large extent focused on the extraction of correlational patterns. Econometricians have long criticized the inability of machine learning algorithms to capture causal relationships between variables of interest. Against this background, the third part of the course examines recent developments in the scope of causal machine learning. Considering the example of decision models in marketing, the course briefly revisits some fundamentals related to causal inference and elaborates on selected causal machine learning algorithms such as causal forests (Athey & Imbens, 2019) or the x-learner (Künzel et al., 2019). Another typical critic machine learners are facing concerns a lack of interpretability. Machine learning models are often considered black boxes. However, recent research has proposed a set of explanation methods for understanding and diagnosing such models. Acknowledging the cruciality of explaining model-based recommendations in many applications of machine learning, Part III of the course will look into the field of interpretable machine learning and equip students with a solid understanding of the options to explain model predictions. Furthermore, quality features, the typical "obstacles" in the planning phase of the research project as well as the collection and analysis of the data are discussed. In addition, the correct way of "writing down" the methodology - especially from the reviewer and the editor perspective - is addressed. In order to learn about these contents, the participants will analyze published "best practices" and apply the findings to their own research projects. Overall, the participants should acquire a basic knowledge of the methodology, in order to independently carry out a qualitative and quantitative content analysis, which meets the international scientific standards.
Period30 Aug 20217 Sep 2021
Event typeCourse
LocationBerlin, Germany, BerlinShow on map
Degree of RecognitionInternational