Milad Shahvaroughi Farahani - MSc in Financial Engineering, Department of Financial Management, Khatam University, Tehran, Iran
Seyed Hossein Razavi Haji Agha - Assistant Professor of Management, Khatam University, Tehran, Iran
Saeed Rahimian - Assistant Professor of Financial Management, Khatam University, Tehran, Iran
Babak Majidi - Assistant Professor of Computer Engineering, Khatam University, Tehran, Iran

Stock price prediction is a very important topic for investors and corporations because through its forecasting, they can increase their profits and raise their capital. Therefore, investigation of the stock price movements and exact determination of the future of the stocks is at the center of attention for financial researchers. However, stock price movements follow multiple complicated factors which result in difficulty of forecasting the exact stock price movements. Consequently, more exact and innovative models and methods to prediction of stock price are developed in recent years. The aim of this study is to evaluate the efficiency of using technical indicators such as closing price, lowest price, highest price, exponential moving average, etc. in prediction of stock prices. For more exact examination of the relationship between technical indicators and stock prices in the considered time intervals, we used Artificial Neural Network (ANN), Genetic Algorithm (GA), Particle Swarm Optimization (PSO) and Harmony Search (HS) algorithms as metaheuristic methods for stock price prediction. The GA is used for selection of the best optimization indicators. Beside selection of optimized indicators, we used PSO and HS to training the neural network, minimizing the network error, obtaining optimized weights and the best number of hidden layer simultaneously. In order to compare the proposed models performance and to choose the best model according to the amount of error, we used eight estimation criteria for error assessment. In experimental results show that a hybrid ANN-HS algorithm has the best performance. Finally, we used Run tests as a non-parametric test for testing the EMH in a weak form.

Technical Indicators, Artificial Neural Network, Genetic Algorithm, Harmony Search, particle Swarm Optimization algorithm