Parameter Optimization for Trading Algorithms of Technical Agents

The focus of this study is on the analysis of indicators commonly used by technical traders, also called technical agents in this article, with the aim of developing a multi-agent model for the financial market in future research. Accordingly, we study different approaches to optimizing the parameters of a trading strategy based on technical analysis, namely Genetic Algorithm and Bayesian Optimization. The trading strategies are built based on two indicators, Relative Strength Index (RSI) and Bollinger Band (BOLL). The experiment is performed in the Dow Jones stock market where 15-minutes price data is used to backtest trading strategies for short-term purposes. The results show that the parameter optimization method using Bayesian Optimization gives better results in terms of return and execution time than the Genetic Algorithm approach in all periods. The optimized strategies from Bayesian Optimization approach can generate higher cumulative returns than their typical form and Buy and Hold strategy. Among the indicators being studied, RSI can beat BOLL with a more profitable and stable optimal strategy.