An interpretable uni-nullneuron-based evolving neuro-fuzzy network acting to identify Dry Beans

• Published in: IEEE International Fuzzy Systems conference proceedings pp. 1 - 9
• Main Authors: de Campos Souza, Paulo Vitor, Lughofer, Edwin
• Format: Conference Proceeding
• Language: English
• Published: IEEE 18.07.2022
• Subjects: Adaptation models; Dry Bean identification; Evolving neuro-fuzzy approaches; Feature extraction; Fuzzy neural networks; fuzzy rules; Interpretability; Knowledge engineering; Streaming media; Training; Transforms
• ISSN: 1558-4739
• DOI: 10.1109/FUZZ-IEEE55066.2022.9882789

Evolving systems are models which are able to act dynamically in adaptive open-loop manner for solving data stream modeling problems within different application areas. Their parametric adaptability for architectural constructions of models allows them to flexibly solve issues of the most varied natures. Data mining problems in the agriculture area are the target of recent researches, mainly through extracting the basic characteristics from images to transform them into feature data that can be processed by machine learning approaches. This work aims to address the problem of identifying dry beans with interpretable models and results. For this purpose, an evolving neuro-fuzzy network based on uni-nullneurons, capable of extracting knowledge about a data set through if-then rules, was used in this paper. The data set used in this work was the subject of several approaches in the literature, and the results obtained (reaching about 98.18% classification accuracy) prove that the evolving neuro-fuzzy network used in this paper can identify dry seed beans with a high degree of precision while allowing the interpretation and dissemination of knowledge about their correct identification.