Stock Market Forecasting Using S-System Model

Abstract

In this paper, S-system model is first presented to forecast stock market index. An improved additive tree model named restricted additive tree (RAT) is proposed to represent S-system model. A hybrid evolutionary algorithm based on structure-based evolutionary method and cuckoo search (CS) is used to evolve the structure and parameters of RAT model. Shanghai stock exchange composite index is used as example to evaluate the performances of S-system model. Results show that S-system model outperforms other traditional models, including neural network, wavelet neural network, flexible neural tree and ordinary differential equation.