This paper proposes a method to discover time series patterns predictive of the occurrence of specified events. The process of technique realization is mainly composed of four steps: specifying interest function and parameters; searching ODs, clustering ODs for candidate patterns and identifying patterns. A simulation study is conducted as verification. And an application study in the stock market is introduced to demonstrate its performance.
Published in:
Natural Computation, 2008. ICNC '08. Fourth International Conference on
(Volume:7
)
Date of Conference: 18-20 Oct. 2008