International Journal of Hybrid Intelligent Systems

International Journal of Hybrid Intelligent Systems

Volume 1, No. 2 (2004)

Novel Time Series Analysis and Prediction of Stock Trading Using Fractal Theory and Time-Delayed Neural Networks

Fuminori Yakuwa, Mika Yoneyama and Yasuhiko Dote

Abstract. In this paper the Nikkei stock prices over 1500 days, from July 1996 to October 2002, are analyzed and predicted using a Hurst exponent (H), a fractal dimension (D), and an autocorrelation coefficient (C). They are H=0.6699, D= 2-H = 1.3301 and C= 0.26558 over three days. In order to extract the knowledge, decision making rules comprehensible by humans using the features are derived by rough set theory. Then this obtained knowledge is embedded into the structure of our developed time delayed neural network (Shafique and Dote 2000). It is a back propagation neural network with a FIR (Finite Impulse Response) filter of the second order plugged into each time delayed input node. It is confirmed that the obtained prediction accuracy is much higher than that obtained by a back propagation-type forward neural network without filters for the short–term. Therefore, this predictor is one of hybrid intelligent systems which is expected to be a promising approach in the future.

Keywords: time series analysis, fractal theory, Hurst exponent, data mining, stock trading.