Application of Patter Recognition in Financial Engineering

Wednesday 30 November, at 11.30, in room X, Dr Aistis Raudys from Institute of Mathematics and Informatics, Lithuania, will give a
talk on "Application of Patter Recognition in Financial Engineering".

Abstract

I will overview the research I am doing at the moment. Most of my research activities are related to trading financial assets. So first, I will overview an automated trading – the increasingly popular topic in finance. Later I will discuss the market liquidity and activity forecasting problemi. These types of forecasts are particularly interesting for big order execution that can impact the market. Precise liquidity forecasting allows us to distribute our orders in a better fashion and save on transactions costs. In one of my works I analyzed market liquidity seasonality and employed several pattern
recognition techniques.
Synthetic time series history is a very interesting topic for me. There are a lot of very liquid exchange trading funds (ETF) that can be used for automated trading but the history available is too short to back test the system reliably. Hence the initiative is to try to reconstruct the missing history. In the current research study we used 7 methods for this task and found that expectation maximization algorithm is showing the best results.
Another topic of my research is portfolio construction for automated trading systems. The classical Markowitz mean-variance portfolio optimization method is expecting time series to be normally distributed. The product of automated trading systems is far from Gaussian distribution. Therefore we examined several ways of solving this problem. One approach is to use covariance matrix normalization. Second one is to use empirical approach and create a portfolio with no assumptions about the distribution of the underlying data. And finally one may employ multi agent approach to this problem by creating a number of diverse portfolios and using the best one.

Dr. Aistis Raudys received his PhD from the Institute of Mathematics and Informatics, Lithuania. The topic of his thesis was feature extraction from multidimensional data. Currently he works at Vilnius University Faculty of Mathematics and Informatics. He is a lecture for computer based trading technologies to undergraduate students. Previously Aistis worked as a researcher and also as a developer in various software companies. He collaborated with number of top tier banks including Deutsche Bank, Société Générale and BNP Paribas. Aistis’s research interests are in machine learning for financial engineering and automated trading. He is the author of 15 publications and scientific works.