Optimizing the Prediction Model of Stock Price in Pharmaceutical Companies Using Multiple Objective Particle Swarm Optimization Algorithm (MOPSO)

The purpose of the research was to optimize the prediction model of stock price in pharmaceutical companies using meta-heuristic algorithm. In this research, optimizing the stock portfolio has been done in two separate phases. The first phase is related to predicting the stock future price based on the past stock information, in which predicting the stock price was done using an artificial neural network. The neural network used in the research was Multilayer Perceptron (MLP) using the back propagation of error algorithm. After predicting the stock price with a neural network, the predicted price data was used to optimize the stock portfolio in the second phase. In the second phase, a multi-objective genetic algorithm was used to optimize the portfolio so the optimal weights are assigned to the stock and the optimal stock portfolio was developed. Having a regression model, the relevant genetic algorithm was designed using MATLAB software. The results showed that the stock portfolio developed by MOPSO algorithm has a better performance under all four risks criteria except the conditional value-at-risk criterion than the algorithms used in the compared article. In all models except the mean-conditional value-at-risk model, the stock portfolios developed by the MOPSO algorithm used in the research have more and more appropriate efficiency. In addition, the results showed that MOPSO algorithm is of good performance at developing and optimizing the stock portfolio and better than other algorithms. Therefore, it can be said that using meta-heuristic MOPSO algorithm used in the research is effective for optimizing stock portfolio.