正则化训练的神经网络与粗集理论相结合的股票时间序列数据挖掘技术

论文提出将正则化神经网络与粗集理论相结合应用于股票时间序列数据库的数据挖掘.首先对时间序列数据库进行预处理,除去高频干扰信号,然后将股票时间序列数据按照收盘价的变化趋势分割成一系列静态模式,每种模式代表股票价格的一种行为趋势(上涨或下跌),把决定各种模式的相关属性组成一系列信息,形成一个适用于粗集方法的信息表.然后使用正则神经网络对信息表进行学习,用粗集理论从正则神经网络所存储的知识中抽取规则,得到的规则可以用于预测时间序列在未来的行为。该方法融合了正则神经网络优良的泛化性能和粗集理论的规则生成能力,实验表明,该方法预测效果比较准确。

Stock Market Time Series Data Mining Based on Regularized Neural Network and Rough Set

This paper presents a new method of stock market time series data mining. It combines regularized neural network with rough set. The process includes preprocessing of time series and data mining. The preprocessing cleans and filters time series. Then, the time series are partitionel into a series of static patterns, which is based on the trend (i.e., increasing or decreasing) of closing price. The most important predicting attributes identified from every model form an information table. The regularized neural network is used to learn and predict the data. Rough set can extract rule knowledge in the neural network, which can be used to predict the time series' behavior in the future. This method combines the generalization faculty of regularized neural network and the rule reduction capability of rough set. The experimental results demonstrate the effectiveness of the algorithm.