A Functional Data Classifier Based on Bonferroni Mean Fuzzy K-Nearest Centroid Neighbor

Instead of representing data as vectors of fixed dimensions, functional data is represented as functions or curves. Functional data analysis can reduce the dimensionality by feature extraction to process and analyze high-dimensional, dynamic, and complex data. Compared to traditional classification methods, functional data classification provides a higher level of data information extraction by considering the entire function or curve rather than individual data points. This article proposes a functional data classifier based on Bonferroni Mean fuzzy K-nearest Centroid Neighbor. Specifically, the features of the data are first extracted using functional data analysis to obtain the basis coefficients, which are used as inputs to the classifier. Then, the Bonferroni Mean based Fuzzy K-Nearest Centroid Neighbor classifier is used to compute the fuzzy membership of the unclassified samples in each category and classify the samples according to the membership of different categories. We tested the performance of the proposed classifier using eight different real functional datasets. Based on the experimental results, compared with the other six functional data classifiers, the proposed method has the best overall classification performance on all datasets.