Optimizing filter rule parameters with genetic algorithm and stock selection with artificial neural networks for an improved trading: The case of Borsa Istanbul

摘要

Filter rule along with other trading algorithms is used to identify potentially profitable trading points in stock markets. In this study, the scope of the filter rule has been expanded to include different moving average types. The filter rule parameters that will provide the highest return for each of the stocks listed in Borsa Istanbul have been optimized by using genetic algorithm. A number of 357 stocks traded in Borsa Istanbul is included in the dataset of the study between 06-07-2012 and 31-03-2020 period. To improve the poor performance in out-of-sample sets of optimal rules, the stock selection process was performed by means of artificial neural networks. The artificial neural network model predicts the performance of the stock in the test set by using the performance values in the training set. Results indicate that the returns of the selected stocks are significantly higher than the returns of the buy and hold strategy. Parameter optimization of filter rule with genetic algorithms and stock selection with the artificial neural networks can be used as a decision support system for investors, where they can make a profit above the market return. When only the genetic algorithm results are taken into account, it can be stated that Borsa Istanbul is a weak form efficient market. However, selecting the stocks with the assistance of artificial neural networks made it possible to obtain excess returns over the market.