Enhancing Existing Stockmarket Trading Strategies Using Artificial Neural Networks: A Case Study

Abstract

Developing financially viable stockmarket trading systems is a difficult, yet reasonably well understood process. Once an initial trading system has been built, the desire usually turns to finding ways to improve the system. Typically, this is done by adding and subtracting if-then style rules, which act as filters to the initial buy/sell signal. Each time a new set of rules are added, the system is retested, and, dependant on the effect of the added rules, they may be included into the system. Naturally, this style of data snooping leads to a curve-fitting approach, and the resultant system may not continue to perform well out-of-sample. The authors promote a different approach, using artificial neural networks, and following their previously published methodology, they demonstrate their approach using an existing medium-term trading strategy as an example.