Creating trading rules using hybrid of genetic algorithms and simulated annealing to confirm buy/sell signals
Trading rules are widely used for market assessment and timing by investors, however due to simplicity of these rules, they often behave poorly in some market conditions. To alleviate this weakness they need to be used as a combination. One major difficulty for combination of trading rules is searching for values of their parameters. This study presents a new hybrid of genetic algorithms and simulated annealing (GASA) used to optimize the combination of common indicators to confirm buy and sell signals in the Foreign Exchange (Forex). Genetic algorithms and simulated annealing are powerful optimization methods with complementary strengths and weaknesses. This thesis proposes an algorithm that repeatedly adds one indicator at a time to be optimized by a genetic algorithm and then applies a simulated annealing step, which decreases the search range of all the parameters of the indicators. Previous research has usually optimized a small number of indicators or optimized each indicator independently. Optimizing all the indicators together results in a search space that is combinatorial; the new algorithm attempts to avoid a large search space by adding indicators incrementally. The results are compared to Buy & Hold, Optimized Moving Average, and Optimized MACD. The new algorithm shows some remarkable improvement in comparison with these strategies.