UNIverse - Public Research Portal
Project cover

D-132 Automated Trading System 8211; Design and Implementation of a robust and adaptive regime-switching model of recurrent reinforcement learning for algorithmic trading

Research Project
 | 
01.01.2009
 - 31.12.2009

The primary objective of this project is to implement an algorithmic trading model that shows a certain level of consistency in its performance, i.e. it should do well over a relatively long period and across different types of time series data. Although consistency is a key issue in algorithmic trading, it has often been neglected. This research project aims at addressing this problem by making use of advances in both econometrics and artificial intelligence. We propose a novel approach based on regime-switching and recurrent reinforcement learning (RRL) that is completely data-driven. It requires the design and implementation of a comprehensive system that consists of powerful data processing techniques smartly linked to a quick and efficient learning mechanism. Such a system can give useful insight into market microstructure in addition to fulfilling its fundamental aim of generating intelligent trade signals.

Members (2)

Profile Photo

Dietmar Maringer

Principal Investigator
MALE avatar

Tikesh Ramtohul

Project Member