Evolving fuzzy predictor with multivariable Gaussian participatory learning and multi-innovations recursive weighted least squares: eFMI

This paper proposes a new evolving fuzzy model constructed with an unsupervised recursive clustering algorithm with participatory learning and multivariable Gaussian membership functions. The proposed model, called evolving fuzzy with multivariable Gaussian participatory learning and multi-innovations recursive weighted least squares, uses first-order Takagi-Sugeno functional rules. The rules are extracted by the clustering algorithm that can add a new cluster, delete, merge or update existing clusters. The clusters are created using a compatibility measure and an arousal mechanism. The compatibility measure is computed by Euclidian or Mahalanobis distance according to the cluster’s number of samples. The clusters exclusion method combines age and population to exclude inactive clusters. Redundant clusters are merged if there is a noticeable overlap between two clusters. The consequent parameters are updated by a multi-innovations weighted least squares recursive algorithm. The performance of the eFMI is evaluated and compared with alternative state-of-the-art evolving models in times series forecasting and non-linear system identification problems. Computational experiments and comparisons suggest the proposed model performs better or comparable than the alternative evolving models.