• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>Third International Conference on Neural Networks in the Capital Markets Vol.2 London, England 11-13 October 95 >Stock Price Preciction Using an Integrated Pattern RecognitionParbdigm
【24h】

Stock Price Preciction Using an Integrated Pattern RecognitionParbdigm

机译：使用集成模式识别范例的股价预测

获取原文
获取原文并翻译 | 示例
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Pattern recognition and prediction have a common goal of unraveling complex, but low-order deterministic system dyanmcis with a set of features and decision statistics. Artificial neural networks (ANNs), extended Kalman flterxs, nonlinear filters, and hidden Markov models have attracted a lot of attention in recent years as aemans of modeling chaotic and financial data. The key underlying assumption is that although data may look random at first glance, there must be a low dimensional deterministic structure which can be exploited to identify hidden patterns and to facilitate accurate forecasting. We apply an integrated pattern recognition apradigm to predition of stock prices. The integrated pattern recognition paradigm consists of low dimensional data characterization, feature extraction, feature optimization, and mapping the classifier structure to the underlying "good" feature distribution. We divide the stock price preiction problem into two subsets: time series prediction (fine quantization) and predicting the future stock price in a binary fomat (coarese quantization). We quantify prediction performance as a function of feature dimension, observation duration To, prediction tiem Tp, and classifier topologies to illustrate the improtance of the integrated approach in extracting maximum prediction performance, Furthermore, we extend the integrated pattern recognition paradigm to time series prediction. We were able to achieve over 90
机译：模式识别和预测的共同目标是揭示复杂的，但具有一组功能和决策统计信息的低阶确定性系统。近年来，随着对混沌数据和财务数据进行建模，人工神经网络（ANN），扩展的Kalman flterx，非线性滤波器和隐马尔可夫模型吸引了众多关注。关键的基本假设是，尽管乍一看数据可能看起来是随机的，但是必须有一个低维确定性结构，可以利用它来识别隐藏的模式并促进准确的预测。我们将集成模式识别方法应用到股票价格的掠夺中。集成的模式识别范例包括低维数据表征，特征提取，特征优化以及将分类器结构映射到基础“良好”特征分布。我们将股票价格预测问题分为两个子集：时间序列预测（精细量化）和预测二元格式的未来股票价格（粗略量化）。我们将预测性能量化为特征维，观察持续时间To，预测联系Tp和分类器拓扑的函数，以说明集成方法在提取最大预测性能方面的重要性。此外，我们将集成模式识别范式扩展到时间序列预测。我们能够达到90多个

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号