基于特征提取的金融时间序列形态挖掘

以金融时间序列的上升三角形形态为例给出了金融时间序列挖掘与特征提取的方法,在挖掘过程中充分利用挖掘者的经验背景知识,提出了一种基于特征提取的金融时间序列形态挖掘算法。该算法首先对金融时间序列进行趋势化预处理,将曲线转化为折线表示,形成形态的趋势片断,然后再从这些趋势片断中提取出相关形态的属性特征。它提高了序列形态特征提取的有效性,使搜索空间大为减小。

Pattern Mining on Financial Time Series Based on Extraction of Features

This paper takes the example of up sloping triangles pattern and offers the methods to pattern mining on financial time series and to identify their features. During the course of research, it makes full use of the experiences and background knowledge of the researcher, and puts forward an exploration and calculation method for the pattern mining into financial time series based on extraction of features. First of all, the calculation method processes the financial time series in order to show them as trends, and then extracts the features of relevant patterns from these segments of trends. This enhances the efficiency of extraction of the features of financial time series patterns, and reduces the search space significantly.