Feature Selection Algorithm for Fuzzy Relation Decision Systems

Feature selection is an important aspect of data dimensionality reduction research and also a hot topic in machine learning. Since Boolean values cannot fully express information, the Boolean values in existing relation matrices are not sufficient to describe the connections and distinctions between objects in relation decision systems. Based on the relation decision system, this paper proposes the concept of the fuzzy relation decision system by using fuzzy sets to describe the relationships between objects, thereby enhancing the robustness and generalization ability of the model and further extending the concept of relation decision systems. Moreover, for the specific feature selection problem in fuzzy relation decision systems, the paper proposes a definition of attribute importance and its calculation method, thereby further expanding the application scope of feature selection. Finally, by using attribute importance, a corresponding feature selection algorithm is proposed, providing a new method for solving engineering application problems.