Effects of Attribute Selection in Multiobjective Fuzzy Genetics-based Machine Learning on Fish Habitat Assessment

Abstract

Urbanization, field development, and water pollution have deteriorated aquatic environments. Therefore, the conservation and restoration of ecosystems are urgent issues in preserving biodiversity. To address these issues, it is important to model the environmental preferences of aquatic organisms and evaluate their habitats. Multiobjective Fuzzy Genetics-based Machine Learning (MoFGBML) can generate interpretable models composed of fuzzy if-then rules to assess the environmental preferences of aquatic organisms. The number of attributes in training data significantly affects the interpretability and classification accuracy of fuzzy models. In this paper, we evaluate the effects of attribute selection on the model performance of MoFGBML and the resulting fuzzy if-then rules that explain environmental preferences of a species.