Bonferroni mean based fuzzy K-nearest neighbor for classification with dimensional reduction in multiclass data

In this study, we demonstrated a new method of multiclass data classification by combining Fuzzy K-Nearest Neighbor (FKNN) and Latent Discriminant Analysis (LDA) as dimension reduction techniques. In addition, we also introduced the Bonferroni Mean approach as part of the Fuzzy KNN algorithm. The purpose of this study is to evaluate and improve classification performance in multiclass data. The results of this study show that by combining the Fuzzy KNN, LDA and Bonferroni Mean techniques, this method is able to produce an precision, recall, F1, accuracy, and balanced accuracy, with an average of 99.3% and 99.4%. The implementation of dimension reduction techniques using LDA helps to solve complexity and overfitting problems in multiclass data. Meanwhile, the Bonferroni Mean approach helps improve the robustness of the Fuzzy KNN algorithm. The study suggests that this new method can be used as an effective and efficient solution for multiclass data classification.