Machine-Learning Fully Automated FX Trading System for Superior Returns with Risk Cybernetics Artificial Intelligence Framework

Abstract

The team at Finamatrix has shown through real trading accounts that it is possible to produce superior returns i.e. 100% on a weekly to monthly basis with manual trading. However, the sustainability of the superior returns are subject to the limitations of human thinking processes that are prone to mistakes due to emotions, cognitive errors, etc. The human brain of an experienced trader is able to harness all the data available and convert this data into useful information in a very short time for profit but this is not a perfect process. We attempt to develop the holy grail of trading systems -- a completely self-learning and self-enhancing trading database system that learns from trader experience and is capable of a suite of functions, including not only recalling and implementing strategies when suitable conditions arise but also capable of adaptation and self-improvement. The system is able to execute continuous iterations and is able to convert errors into learning experiences so as to re-program previous bad trades to create good trades. The system also features learning from good trades and is able to modify and re-input the parameters so as to offer relatively more appropriate trading strategies that are able to produce better returns. These functions are potentially made possible through the utilization of genetic-optimization neural-networks in a risk cybernetics framework.

Keywords: artificial intelligence, automated trading systems, algorithmic trading, cybernetics, machine-learning

JEL Classification: C45