Knowledge engineering

Pattern recognition for artificial financial market data

A. O. Glekin

Institute for Systems Analysis of Russian Academy of Sciences

Abstract: In this paper a learning algorithm is proposed for prediction a behavior of the financial time series when new information affecting the market. The algorithm is designed for nonlinear markov model. The prediction problem is formulated as classification problem. Bagging is used to build efficient combined classifiers. Minimum Hamming distance classification method with genetic learning algorithm is used as individual classifiers. Experiments on real financial time series show that proposed algorithm has a fairly high generalization performance.

Keywords: multi-agents model, artificial financial market, classification, bagging, genetic algorithms.

Bibliographic databases:

•