SEEM A Structural Enhancement of evolving method for fuzzy system

Al-Marhabi Zaid, Yahya Al-Ashmoery, Hisham Haider Yusef Sa’ad (*), Taher Alreshahi, Mofaq Albaraq, Adnan Haider

  • Al-Marhabi Zaid Al-Razi University
  • Yahya Al-Ashmoery Al-Razi University
  • Taher Alreshahi Hodeidah University,
  • Hisham (*) Haider Yusef Sa’ad Al-Razi University
  • Mofaq Albaraq National University
  • Adnan Haider Al-Razi University
الكلمات المفتاحية: Evolving methods, incremental learning, function approximation


Determining the optimal fuzzy terms, which lead to the minimum global error, remains a challenge in the fuzzy methods. Another challenge is that the parameters of the consequent section are usually selected individually thus resulting insufficient fuzzy systems. In this paper, A Structural Enhancement of evolving method (SEEM) has been proposed to solve such problems. SEEM has been developed evolutionarily based on incremental partitioning learning. SEEM begins with an initial fuzzy system that has double fuzzy terms for the antecedent part. Then, to create a more accurate fuzzy system, it keeps improving by identifying the ideal input fuzzy term and ideal consequent parameter. There are two steps involved in determining the antecedent component and the consequent parameters. By detecting the distinction points (extremum and inflection points) using the gradient descent approach, it first identifies fuzzy terms and the consequent parameters. This continues until all of the fuzzy terms and consequent parameters are obtained. The second step is identifying the ideal new fuzzy terms that produce the global best result. This model utilizes the gradient descent estimator to obtain the optimum consequent parameters. As a result, SEEM produces sufficient fuzzy systems that have fewer number of fuzzy rules with high accuracy.