An investigation of artificial neural networks applied to pairs trading in South Africa

Abstract

The use of artificial intelligence models could enhance the process by which decisions are made for trading in the financial markets. In this project the aim was to define, create, apply and analyse the results of an artificial neural network based solution for so-called pairs trading. A literature study is presented where a brief summary is given of the current body of knowl- edge with regards to artificial neural networks and their application in finance. Additionally an introduction of the software packages used in this project is given. In a separate chapter a short introduction on trading strategies and pairs trading, specifically, is included. The design of the artificial neural network is discussed in detail with both training and execu- tion results from experiments critically examined. Finally the question ‘Are artificial neural networks a viable tool applied to pairs trading in the South African market?’ is answered based on the empirical evidence.