Hybrid metaheuristic artificial neural networks for stock price prediction considering efficient market hypothesis

Investigating stock price trends and determining future stock prices have become focal points for researchers within the finance sector. However, predicting stock price trends is a complex task due to the multitude of influencing factors. Consequently, there has been a growing interest in developing more precise and heuristic models and methods for stock price prediction in recent years. This study aims to assess the effectiveness of technical indicators for stock price prediction, including closing price, lowest price, highest price, and the exponential moving average method. To thoroughly analyze the relationship between these technical indicators and stock prices over predefined time intervals, we employ an Artificial Neural Network (ANN). This ANN is optimized using a combination of Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Harmony Search (HS) algorithms as meta-heuristic techniques for enhancing stock price prediction. The GA is employed for selecting the most suitable optimization indicators. In addition to indicator selection, PSO and HS are utilized to fine-tune the Neural Network (NN), minimizing network errors and optimizing weights and the number of hidden layers simultaneously. We employ eight estimation criteria for error assessment to evaluate the proposed model's performance and select the best model based on error criteria. An innovative aspect of this research involves testing market efficiency and identifying the most significant companies in Iran as the statistical population. The experimental results clearly indicate that a hybrid ANN-HS algorithm outperforms other algorithms regarding stock price prediction accuracy. Finally, we conduct run tests, a non-parametric test, to evaluate the Efficient Market Hypothesis (EMH) in its weak form.