Course Prerequisites: No prerequisites, but successful participation in at least one of the following statistics/data science courses isrecommended: Advanced Marketing Modeling, Selected Topics in Statistical and Machine Learning, Business Analytics & Data Science, Advanced Data Analytics for Management Support. In addition, students should be proficient in Python and have a good understanding of statistics, probability theory and linear algebra.
Description and Objectives: This course is designed for master students from quantitative fields such as marketing, economics, statistics and computer science in their last year of study. It prepares students for solving real-world marketing problems using modern quantitative methods and is a good preparation for a machine learning/data science job in marketing or a PhD in quantitative marketing.
To this end, the course first reviews theoretical foundations in marketing, statistics, probability theory and computer science that are required to understand, apply and customize complex statistical models. The course will then focus on formalizing marketing decisions as machine learning problems and equips students with the necessary tools to efficiently implement machine learning models and pipelines. After completing this course participants will be able to judge how modern machine learning methods complement (or even replace) traditional statistical methods for data analysis and decision-making.
The course content complements existing courses in that it reviews the theoretical foundations taught in statistics and computer science programs and then shows how to implement machine learning approaches to important marketing questions.
Course material will be made available in the Moodle system of the Humboldt-University Berlin.
Grades are solely based on the final (written) exam. Successful participation in the home assignments serves as a preparation for the exam and is not mandatory for admission. Reference solutions to the home assignments will be discussed in the exercises.
All lectures and exercises are based on Python. The students will use popular data and machine learning libraries including (among others) numpy, scipy, pandas, scikit-learn, pytorch, lightgbm and statsmodels. Students can run Python via jupyterlab or in their preferred IDE. It is strongly recommended to manage libraries using virtualenv. The sample code is tested on Unix and students are encouraged to use a Unix OS (Mac OS X or Linux). Support for Windows is not guaranteed.