Hybridizing extreme learning machine and bio-inspired computing approaches for improved stock market forecasting

Abstract

Under today's economic conditions, developing more robust and realistic forecasting methods is needed to make investments more profitable and secure. However, understanding the structure of the stock markets is very difficult because of the dynamic and non-stationary data. In this context, bio-inspired computing approaches including evolutionary computation and swarm intelligence can be used to make more accurate calculations and forecasting results. This paper improved Extreme Learning Machine (ELM) using Genetic Algorithm (GA), Differential Evolution (DE) as a two evolutionary computation methods, and Particle Swarm Optimization (PSO) and Weighted Superposition Attraction (WSA) as a two swarm intelligence methods for stock market forecasting in Turkey. The results of this study show that proposed methods can be successfully used in any real-time stock market forecasting because of the noteworthy improvement in forecasting accuracy.