Study on Financial Data Hybrid Clustering Based on Stock Trading Rule
We have built a SOM&K-means based trading rules system on financial markets at intraday trading frequencies and conduct empirical research on the 5 min Hushen 300 Index of China. We also have obtained the result that K-means clustering after the train
- PDF / 127,238 Bytes
- 7 Pages / 439.37 x 666.142 pts Page_size
- 114 Downloads / 177 Views
Study on Financial Data Hybrid Clustering Based on Stock Trading Rule Jian Huang
Abstract We have built a SOM&K-means based trading rules system on financial markets at intraday trading frequencies and conduct empirical research on the 5 min Hushen 300 Index of China. We also have obtained the result that K-means clustering after the training of the SOM can categorize successfully. The new proposed SOM&K-means based trading rules can provides purchase decision to investors easily. The results support the evidences that SOM&K-means based trading rules system will discover intraday patterns and help stock investors gain more returns for the stock investors. Keywords Stock trading rule • K-means • Financial data hybrid clustering
300.1
Introduction
Making correct trade decision on time is a challenge for investors in the stock market. By literature, a number of scholars have researched on this topic, from traditional technical analysis based trading rules to recent machine learning based stock market forecasting. In finance, technical analysis is a security analysis discipline for making trading rules by study past market data, price and volume. Random walk [1], moving average [2], and trading range break-out rule [3] are the well known technical analysis methods. In recent years, machine learning method has been introduced to handle financial problem, which include decision tree, clustering, genetic programming, and artificial neural networks and so on [4, 5]. Almost every aspect of financial markets has been investigated by artificial neural networks [4]. But most of the published scientific papers investigated into the
J. Huang (*) Computer Science and Information Technology School, Zhejiang Wanli University, 315100 Ningbo, China e-mail: [email protected] S. Zhong (ed.), Proceedings of the 2012 International Conference on Cybernetics 2351 and Informatics, Lecture Notes in Electrical Engineering 163, DOI 10.1007/978-1-4614-3872-4_300, # Springer Science+Business Media New York 2014
2352
J. Huang
developed countries’ stock markets based on the low frequency financial data, which are observed yearly, monthly, or daily. There are a few literature focuses on China’s stock market, especially the Shanghai and Shenzhen stock markets [6, 7].
300.2
Preliminary
SOM algorithm. In 1982, Kohonen proposed self-organizing maps (SOM) network [8]. SOM is an unsupervised, self-organizing, adaptive network, which has been widely used in various fields for data analysis. SOM’s basic idea is make the highdimensional data project to a two dimensional space and achieve data clustering through training network. The approach will facilitate reducing data dimensions and achieve the data visualization. i denotes the weight vector Let xk 2 ave (1 + band) then set signal of St + 1 to buy (+1), Where band is a non-negative threshold. If c < ave (1band) then set signal of St + 1 to sell (1) Set signal of St + 1 to standby (0) in other cases.
(e) Repeat (d) for all the test samples. Figure 300.2 shows the process of SOM&K-
Data Loading...
