Detecting Changes of State in Heterogeneous Dynamic Processes with Time-dependent Models: A Soft-Computing Approach

J.J. Valdés (Canada)


neuro-fuzzy networks, evolutionary algorithms, probability distributions, heterogeneous multivariate time series, model discovery.


This paper introduces a computational intelligence approach to the problem of detecting internal changes within dynamic processes described by heterogeneous, multivariate time series with imprecise data and missing values. A data mining process oriented to model discovery using a combination of neuro-fuzzy neural networks and genetic algorithms, is combined with the estimation of probability distributions and error functions associated with the set of best discovered models. The analysis of this information allows the identification of changes in the internal structure of the process, associated with the alternation of steady and transient states zones with abnormal behavior instability and other situations. This approach is rather general, and its potential is revealed by experiments with simulated and real world data.

Important Links:

Go Back