EFNN-NullUni: An evolving fuzzy neural network based on null-uninorm

Interpretability in intelligent models becomes a challenge academic research and approaches that facilitate understanding the responses obtained based on artificial intelligence machine learning. This paper presents new logical fuzzy neuron concept of null-uninorm, thus called null-unineuron to compose architecture an evolving neuro-fuzzy model. structural can extract advanced rules allowing AND OR-connections antecedents better interpret understand analyzed problem. three-layer model uses weighted fuzzification approach incremental data partitioning concepts for knowledge extraction through null-unineurons whose training procedure suits classification binary multiclass patterns online way. The weights integrated algorithm belong feature importance levels achieve automatic shrinkage distance calculations along unimportant input directions (features), which turn accounts soft dimension reduction likelihood decrease over-fitting. proposed this was, subject pattern tests, being more efficient compared related (evolving) literature. Finally, experiments various real-world sets proved act simplified way from while providing answers with high degree accuracy problems.