Mining latent sources of causal time series using nonlinear state space modeling

Abstract

Data mining refers to use of new methods for the intelligent analysis of large data sets. This paper applies one of nonlinear state space modeling (NSSM) techniques named nonlinear dynamical factor analysis (NDFA) to mine the latent factors which are the original sources for producing the observations of causal time series. The purpose of mining indirect sources rather than the time series observation is that much better results can be obtained from the latent sources, for example, economics data driven by an «explanatory variables» like inflation, unobserved trends and fluctuations. The effectiveness of NDFA is evaluated by a simulated time series data set. Our empirical study indicates the performance of NDFA is better than the independent component analysis in exploring the latent sources of Taiwan unemployment rate time series. © 2011 Springer Berlin Heidelberg.