MA Seminar – Machine learning in finance

MA Seminar – Machine learning in finance

Prof. Jens Jackwerth

WS 2024/2025

for students in FiMa and Economics

in English

Block seminar: in person January 24, 2025 and January 25, 2025, tba, 09:00h.

First meeting: November 4, 2024, 14:00, F208.

Registration: During the official registration period.

After the enrolment period students cannot drop the seminar any more. Grades will be entered directly into the computer record (no “Schein” available in paper form). No-shows receive a 5.0.

We explore a wide range of machine learning approaches that have recently been applied to finance. We will learn about neural networks and non-parametric regression techniques, as well as about language processing. Particular attention will be placed on understanding the pitfalls of machine learning in finance. All this should give you a handy tool-box for advanced finance problems. 

TOPICS

Types of Machine Learning

1. Supervised, Unsupervised and Reinforcement Learning

2. Support Vector Machine (SVM), Kernel Trick

3. Artificial Neural Networks (ANN)

Applications for Supervised Learning

4. Loans and Insurance Underwriting; Fraud Detection

5. Natural Language Processing (NLP)

Applications for Reinforcement Learning

6. Portfolio Optimization and Risk Management

7. Trading (high frequency, algorithmic trading, ANN, bagging, boosting, meta labelling)

Pitfalls

8. Overfitting Problem and Solutions; Other Pitfalls: Low Signal to Noise Ratio, Weighting of non-IID Data, Structural Breaks

The requirements for this seminar are fourfold and will contribute to the final grade (all contributions are required to be in English):

  • First, and without a grade, you have to submit a short outline by December 1, 2024 (Outlines: via email to office.jackwerth@uni-konstanz.de with the file name such as Last_First_outline.pdf, replies via email). I will look it over and warn you if you are going wrong. This is required but does not count towards the grade.

The preparation of a literature review (at most 10 pages of text, no more than 20 pages in total). I value if you find independent research papers or sources, starting points are www.ssrn.com and www.google.com for looking up authors or to find an electronic copy of a paper you would like to read. Another good source is www.repec.org. Historical data can at time be found on Yahoo Finance. More information about internet search you will find on my website:

https://www.wiwi.uni-konstanz.de/en/jackwerth/teaching/tips-for-internet-research/

  • Weight 50% of the grade.
  • You should send your work, due January 5, 2025, to my secretary Iris Mann (office.jackwerth@uni-konstanz.de):
  •             - an electronic copy of the short paper or the literature review (Word or PDF) with the a file name such as Last_First_paper.pdf
  • I will then mail all papers
  • A discussion of the paper of one of your colleagues (at most 3 pages of text, no more than 6 pages in total). Weight 25% of the grade.
  • You should send your work by January 12, 2025 to office.jackwerth@uni-konstanz.de
  •             - an electronic copy of your discussion (Word of PDF) with the a file name such as Last_First_discussion.pdf
  • I will then put all discussions on my website, too.
  • Participation in person at the seminar on January 24, 2025 and January 25, 2025. We meet in room tba at 09:00 and end the seminar around 18:00. Weight 25% of the grade.

I am looking forward to working with you on this seminar.

Sincerely, Jens Jackwerth