Applying soft computing for forecasting chaotic time series

We present a chaos forecasting system for chaotic time series. After reconstructing the phase space of a chaotic time series, we partition the phase space into some clusters using the fuzzy c-means clustering algorithm. We learn the cluster to which future values will most likely belong. This allows us to make short-term forecasting of the future behavior of a time series by back-propagation network using information based on the cluster. We present an error estimate for this chaos forecasting system and demonstrate its effectiveness by applying it to an example in the logistic map.