Multi-objective optimization approach to stock market technical indicators

Abstract

In this paper the optimization of four technical indicators are analyzed (Exponential Moving Average — EMA; Relative Strength Index — RSI; Moving Average Convergence Divergence — MACD; and Larry Williams %R — %R). In the context of stock markets, technical indicators are widely used by investors to compose trading strategies with the purpose of define entry and exit points of stocks. Although investors want to maximize the profit of their investments with minimum risk, other objectives can be taken into consideration. Despite maximizing profit and minimizing the level of risk, this work also consider minimization of number of trades, what has impact on transaction costs. Multi-objective optimization procedures are used to generate a set of non-dominated solutions (Pareto front) from where the investor can analyze the tradeoff between objectives and choose a solution that best fit its strategy. In this paper an algorithm based on Differential Evolution is used to generate Pareto fronts for each technical indicator. To evaluate the potential excess returns generated using adequate settings for each indicator, the parameters are optimized in a 10 year period daily data from IBOVESPA index. To check whether the optimized parameters can be efficiently used in the future, a period-by-period optimization is performed, where the parameters are optimized in one year period and then tested in the next six month period.