Skip to Main Content
This paper proposes (1) a polygon distribution descriptor and (2) an EC-based similarity measurement for stock market behavior analysis. After learning stock market historical data, a polygon descriptor can capture the dependencies among stock market quantities, such as stock prices, volumes, EPS (earn per share) and so on. By applying the EC-based similarity measurement on polygon descriptors which were trained by stock market data during different periods, the similarity of corresponding stock market behavior can be analyzed. To demonstrate the representation capabilities of the proposed polygon descriptor, Taiwan stock market data from 1986 to 2006 are used. Experimental results show that the polygon descriptor captures the dependencies of stock market quantities, and the similarity measurement shows that the proposed methods capture the changes of market behavior as expected.