Deep Hybrid System of Computational Intelligence for Time Series Prediction

摘要

In this paper the deep hybrid system of computational intelligence for time series forecastingis proposed. As the layer of such hybrid system we propose to use the hybrid generalized additive type-2 fuzzy-wavelet-neural network. Proposed deep stacking forecasting hybrid system of computationalintelligence is enough simple in computational implementation due to parallelizing process ofimplemented computations, has high learning rate convergency due to dismissal from errorsbackpropagation and using the rapid adaptive algorithms for tuning its parameters, has also flexibilitydue to possibility of tuning the activation-membership functions. Proposed system is aimed for solvingthe wide range Data Stream Mining tasks, such as forecasting, identification, emulation, classificationand pattern recognition in online mode.