The main objective is to develop a prototype framework for pricing and risk management using machine learning algorithms and a large variety of heterogeneous and high-volume data, including tick-by-tick quotes of bond prices, market data underlying economic indicators (such as interest rates, foreign exchange rates, inflation rates, and commodity prices) and news feeds. This predictive analytics framework can be used to understand hidden structures in the data and as a test bed for trading risk management with a strong emphasis on back-testing of algorithms in real and artificially simulated environments.
The most important contribution is the development of novel back-testing methodologies of state-of-the-art predictive analytics methods applied to real and simulated environments. The result will potentially have a big impact on trading and risk management practices and the understanding of financial markets complexity in general. A prototype system connected to the data warehouse consisting of variety of machine learning algorithms will be developed to classify data, detect patterns, prediction of prices, patterns, sequences, and cascading effects. Moreover, these algorithms in real and artificially simulated environments will be backtested. For the latter, we will build an agent-based microscopic model with necessary degrees of freedom and evolution rules to mimic a realistic financial market that specialises in trading some selected securities.
Early Stage Resercher working on the project: Ioannis Anagnostou
Supervisors: Dr. Drona Kandhai, ING Bank / drona.kandhai(at)ingbank.com, Prof. Peter Sloot, University of Amsterdam
