Abstract:Epsilon machine is a new algorithm that tries to discover hidden patterns from data. Recently, the scholars in Santefe Institute have already applied it in symbol series successfully, but new problems emerge in traditional time series. A symbolization method transforming the sampling data into symbol series is presented, which implies some information of the expectation and variance. After Causal-State Splitting Reconstruction (CSSR), hundreds of states are lumped in the result, and a new recursion program can pick out the deterministic states very easily. Noise and nonstationarity will stunt the epsilon machine and they are the main problems to be researched in the future.
向馗; 蒋静坪. 一种基于隐含模式发现的时间序列处理算法[J]. 电子与信息学报, 2007, 29(1): 59-62 .
Xiang Kui; Jiang Jing-ping. An Algorithm for Time Series Based on Hidden Pattern Discovery. , 2007, 29(1): 59-62 .