Parameter optimization of oscillator-based technical analysis indicators using evolutionary algorithms on trend-normalized financial data

Technical analysis indicators are widely used technical analysis methods in stock market forecasting. These indicators are obtained by technical analysis and used to produce trade signals for financial instruments. After the researches conducted on financial instruments, easily understandable general trading rules have been introduced for these indicators. However, the results of these general rules are not reliable under the effect of market trend. The performance of the general rules of the indicators are affected by the market trend. As forecasting of the trend is also difficult, the existence of a trend decreases the performance of these general rules. In this study, the parameters of the RSI (Relative Strength Index) and Williams %R indicators are optimized by genetic algorithms and particle swarm optimization for each ETF. Moreover, the same analysis is done under downtrend and uptrend separately to produce two different rules for two different trend type. Finally, the trends of the ETFs are calculated and the normal prices of the ETFs are detrended to remove trend effect from the prices and the same analysis is done on these normalized prices to produce a single trading rule. The resulting rules from the optimization process are tested and the performance of the obtained rules are evaluated. The results show that the rules obtained by detrending the ETF prices are less vulnerable to trend effect.