Generating long-term trading system rules using a genetic algorithm based on analyzing historical data

In current times, trading success depends on choosing a correct strategy. Algorithmic trading is often based on technical analysis — an approach where the values of one or several technical indicators are translated into buy or sell signals. Thus, every trader's main challenge is the choice and use of the most fitting trading rules. In our work, we suggest an evolutionary algorithm for generating and selecting the most fitting trading rules for interday trading, which are presented in the form of binary decision trees. A distinctive feature of this approach is the interpretation of the evaluation of the current state of technical indicators with the help of dynamic ranges that are recalculated on a daily basis. This allows to create long-term trading rules. We demonstrate the effectiveness of this system for the Top-5 stocks of the United States IT sector and discuss the ways to improve it.