Volterra Kernel Constructive Extreme Learning Machine Based on Genetic Algorithms for Time Series Prediction

Time series prediction has become a heavily researched topic in the past several decades because of its broad application scenarios. Although many prediction algorithms have been proposed, few methods are available to generate an optimal prediction model. In this paper, we proposed a novel algorithm based on the Volterra series model and Constructive Selection for Extreme Learning Machine (CS-ELM) to build an effective model for time series prediction. More specifically, we employ genetic algorithms (GAs) to optimize the hidden layer formed by CS-ELM for greater accuracy. The experimental results for several real-world applications show that the proposed algorithm produces better accuracy and generates more effective prediction models than CS-ELM and other classic neural networks (NNs) methods.