| Abstract | This paper explores a computational intelligence approach to the problem of detecting internal changes in time dependent processes described by heterogeneous, multivariate time series with imprecise data and missing values. Processes are approximated by collections of time-dependent nonlinear AR models represented by a special kind of neuro-fuzzy neural networks. Grid and high throughput computing model-mining procedures using neuro-fuzzy networks and genetic algorithms, generate collections of models composed by sets of time lag terms from the time series, as well as prediction functions represented by neuro-fuzzy networks. The composition of the models and their prediction capabilities, allows the identification of changes in the internal structure of the process. These changes are associated with the alternation of steady and transient states, zones with abnormal behavior, instability, and other situations. This approach is general, and its potential is revealed by experiments using paleoclimate and solar data. |
|---|
