AFSKNN: Classification Method Based on K-Nearest Neighbors and Axiomatic Fuzzy Sets

The classification with both semantic description and accuracy performance is an appealing issue in the fields of pattern recognition and data mining. In this study, we make an attempt to design an interpretable classifier AFSKNN based on K-nearest neighbors (KNN) and Axiomatic fuzzy sets (AFS). The algorithm converts the learning data into a semantic description, followed by selecting k-neighborhood of test sample with the aid of the membership grade belonging to the training sample description. Some data sets from UCI machine learning repository are tested for comparing the results with KNN and AFS classifiers. The results demonstrate that the performance of AFS-KNN is satisfied in both interpretability and accuracy. Meanwhile, the classifier indicates a potential of acquiring knowledge by means of semantic description.