Laboratory for Financial and Risk Analytics seminar

The Laboratory for Financial and Risk Analytics seminar series takes place every other Monday at 16:00, at the Faculty of Electrical Engineering and Computing, in the Black Hall. Attendance is free and open, but registration is necessary. 

If you wish to be included in our mailing list for upcoming seminars and other news, please register here:

Upcoming seminars

Monday, March 6, 2023 

Title: Application of deep learning to the problem of determining the elliptic curve rank

Abstract: Generally, in number theory, finding elliptic curves of high rank is of great interest. Such curves have many applications, especially in cryptography. Known algorithms for calculating the rank of a given curve are slow and inefficient, and the current approaches essentially amount to exhaustive search. We propose a new approach to the problem, using a deep neural network to predict the rank of a given curve using other easy to calculate features. This approach can significantly speed up the search for elliptic curves. We also show some new properties of elliptic curves, not known until now. The talk is intended to be possible to follow with no specific number theory knowledge.

Presenter: Domagoj Vlah, Faculty of Electrical Engineering and Computing

Registration link:

Past seminars

Monday, February 6, 2023 

Title: Testing the stability of Croatian insurance companies' portfolios using unsupervised learning

Abstract: The stability of Croatian insurance companies' portfolios is tested by constructing a suitable benchmark and analyzing its diversification levels throughout a historical period of 20 years. Using principal component analysis, the sensitivity of the benchmark to systematic market risk is measured. An inverse relationship between the total diversification of the benchmark and the systematic risk is shown, as well as the existence of unused diversification potential of the benchmark portfolio, in all of the considered periods. The suggested methods are also used to analyze an augmented benchmark contatining instruments for currency, market price and interest rate risk protection, and the results show increased stability with respect to the original benchmark, especially during market crises.

Presenter: Renata Kovačević, Croatia osiguranje

Monday, December 12, 2022 

Title: Electricity prices forecasting

Abstract: The research discusses machine learning approach to electricity prices forecasting in order to assess the battery storage profitability from an out-of-sample perspective. Clustering and Random Forest methods are applied to the German power market data to forecast the day-ahead electricity prices. Discussion topics include model benchmarks, forecast errors and possible links to variables and variable transformations, approaches to combining data clustering and random forest forecasting and clustering metrics. Forecast horizon length and implication for profitability assessment comparisons is also highlighted.

Presenter: Jura Jurčević, Faculty of Economics and Business, University of Zagreb

Monday, November 28, 2022 

Title: Technological challenges in high-frequency trading

Abstract: This talk will discuss the organization of HFT systems, from machine learning models to order submission, and comment on points where statistics or machine learning are used.

Presenter: Igor Sočec, Susquehanna International Group

Monday, November 14, 2022 

Title: Options market making and why options markets are becoming a major mover of markets

Abstract: Using options an individual investor can achieve leverage and a non-linear payout, but a growing options market can have a great impact on the underlying market as well. Market makers dynamically hedge their positions, which can lead to some extreme situations like the gamma squeezes seen last year or more benign, but more tradable situations around option expirations. With growing option volumes and shrinking tenors, options are becoming an important factor in the market and noone can ignore them any more.

Presenter: Goran Dubček, InterCapital Asset Management

Monday, October 17, 2022

Title: Goal-Based Portfolio Allocation – from analytical solutions to reinforcement learning

Abstract: Goal-based investing is wealth management focused on achieving a predefined goal (amount of wealth) within a predefined timeframe. The main characteristic of a goal-based portfolio is the tendency to be fairly risk-exposed in the beginning and decrease risk as the time limit approaches. The primal challenge is to find an optimal allocation strategy that both accumulates the wanted amount of wealth and secures it as the payoff nears. This talk will cover several traditional approaches to dealing with the issue in question. Moreover, the problem of optimising a goal-based portfolio requires sequential decision making and therefore can be implemented in a reinforcement learning framework. The talk will also discuss this machine learning technique as a potential candidate for finding an optimal allocation policy.

Presenter: Tessa Bauman, Laboratory for Financial and Risk Analytics, UNIZG-FER