MIMO evolving participatory learning fuzzy modeling

Evolving participatory learning fuzzy modeling is a flexible and effective method to handle real world complex systems. It is capable to process and learn from streams of data online, and is a natural candidate to find fuzzy rule-based model structures in dynamic environments. This paper extends the evolving participatory learning fuzzy approach for multi-input multi-output - MIMO - processes modeling and suggests the use of subtractive clustering (SC) algorithm to obtain an initial rule base when a priori knowledge is available. SC improves autonomy because it adds learning flexibility. Modeling uses the participatory learning fuzzy clustering algorithm to find rule antecedents, and the recursive MIMO least squares algorithm to estimate the parameters of the linear rules consequents. A novel application concerning modeling the term structure of interest rates and forecasting is also included. Computational results based on the US fixed income market data show that the MIMO evolving participatory learning fuzzy model describes the interest rate behavior accurately, revealing a high potential to forecast complex nonlinear dynamics in uncertain environments.