EXCHANGE RATE FORECASTING USING MULTIPLE CLASSIFIER SYSTEMS

摘要：

Presently, neural networks are widely used in exchange rate forecasting. As know to all, exchange rate may be affected by economical, political, psychological and many other kinds of factors. These factors have different scale and meaning. It is difficult to cope with all of these data with a single classifier. Thus multiple classifier systems are used to deal with this problem in this paper. Firstly, important indicators that fluctuates exchange rate significantly are chosen and categorized so that each category contains only one kind of indicator. Then, classifiers for each category called component classifiers are designed using Radial Basis Function neural networks. Finally, component classifiers are combined to achieve more accurate forecasting. Experimental results demonstrate that multiple classifier systems outperform individual classifier both in sense of NMSE (Norm Mean Squared Error) and DA (Direction Accuracy).

关键词： neural networks multiple classifier systems nonlinear integration economic indicators exchange rate forecasting

作者: Liang Jia Hongwei Hao Xueli Wu

作者单位: School of Information Engineering,University of Science and Technology Beijing,Beijing 100083,China

会议类型: 国际会议

会议名称: 2008年拟人系统国际会议(2008 International Conference on Humanized Systems )(ICHS’08)

会议地点: 北京

会议语种:英文

在线出版日期: 2008-10-18（万方平台首次上网日期，不代表论文的发表时间）