Title:Hybrid symbiotic organisms search feedforward neural network model for stock price prediction

Abstract:The prediction of stock prices is an important task in economics, investment and financial decision-making. It has for several decades, spurred the interest of many researchers to design stock price predictive models. In this paper, the symbiotic organisms search algorithm, a new metaheuristic algorithm is employed as an efficient method for training feedforward neural networks (FFNN). The training process is used to build a better stock price predictive model. The Straits Times Index, Nikkei 225, NASDAQ Composite, S&P 500, and Dow Jones Industrial Average indices were utilized as time series data sets for training and testing proposed predic-tive model. Three evaluation methods namely, Root Mean Squared Error, Mean Absolute Percentage Error and Mean Absolution Deviation are used to compare the results of the implemented model. The computational results obtained revealed that the hybrid Symbiotic Organisms Search Algorithm exhibited outstanding predictive performance when compared to the hybrid Particle Swarm Optimization, Genetic Algorithm, and ARIMA based models. The new model is a promising predictive technique for solving high dimensional nonlinear time series data that are difficult to capture by traditional models.