Identification of stochastic process based on genetic algorithm (人工知能と知識処理)

スポンサーリンク

概要

• 論文の詳細を見る

• This study deals with the problem on how to simultaneously estimate the parameters in a stochastic process, especially under some complicated circumstances. Some previous research suggest using x^2-fitting to estimate these parameters. But, it is certainly difficult to carry a x^2-fitting with several unknown distributional parameters. Here in this study, we suggest estimating these paramters simultaneously by using Genetic Algorithm (GA). At first we explain Tsallis distribution and entropy model related to the Fokker-Planck equation, which is usually used to describe time-space evolution of particles. Since Tsallis distribution can provide dynamical traces of probability density functions (p.d.f) which evolve over different time spans. Different from conventional Brownian motion, Tsallis distribution is evolving as an anomalous diffusion process, and it includes two types distributions, namely, one is a distribution with finite moments, the other is a distribution with infinite moments. Actually there are several parameters to be optimized simultaneously, it is not easy for some simple x^2-fitting to estimate. Thus, we propose to use the GA-based procedure to simultaneously optimize parameters of Tsallis anomalous diffusion process. In our numerical studies, we find that our proposed method works well on tracing the whole evolving picture of returns distribution of the High Frequency Data (HFD) in the stock market.

• 社団法人電子情報通信学会の論文
• 2009-02-23

著者

• Tan Kangrong

  Faculty Of Economics Kurume University

• Tokinaga Shozo

  Faculty Of Economics Kyushu University

• Tan Kangrong

  Faculty of Economics, Kurume University

関連論文

• Identification of stochastic process based on genetic algorithm (人工知能と知識処理)
• Identification of stochastic process based on Genetic Algorithm (知能と複雑系)
• The Design of Multi-Stage Fuzzy Inference Systems with Smaller Number of Rules Based upon the Optimization of Rules by Using the GA (Special Section on Nonlinear Theory and Its Applications)
• Modeling Returns Distribution Based upon Radical Normal Distributions
• Approximating Probability Distribution Function based upon Mixture Distribution Optimized by Genetic Algorithm and its Application to Tail Distribution Analysis using Importance Sampling Method
• Modeling and Analysis of Agent-based Artificial Demand-Supply Market by Using the Genetic Programming and its Applications
• Forecasting the Stock Option Prices by Using the Fractal Geometry and its Application
• Improved Estimation of Tail Distribution of Link Delays by Using Networks Tomography Based on the Genetic Programming and the Importance Sampling

スポンサーリンク