K-Nearest Neighbors algorithm is one of the most used classifiers in terms of simplicity and performance. Although, when a dataset has many outliers or when it is small or unbalanced, KNN doesn't work well. This paper aims to propose a novel classifier, based on K-Nearest Neighbors which calculates the local means of every class using the Power Muirhead Mean operator to overcome alluded issues. We called our new algorithm Power Muirhead Mean K-Nearest Neighbors (PMM-KNN). Eventually, we used five well-known datasets to assess PMM-KNN performance. The research results demonstrate that the PMM-KNN has outperformed three state-of-the-art classification methods in all experiments.
翻译:K- Nearest Mighbors 算法在简单和性能方面是最常用的分类方法之一。 虽然当一个数据集有许多异常值或当它很小或不平衡时, KNN 效果不好 。 本文旨在基于 K- Nearest 邻里bors 提出一个新的分类方法, 该分类方法计算出每个类别的地方手段, 使用Power Muirheadmement Syler 操作员来克服隐蔽的问题 。 我们称我们的新算法“ MuirheadBine K- Nearest Neghirbors (PMM- KNN) ” 。 最后, 我们用五个众所周知的数据集来评估 PMM- KN 的性能。 研究结果显示, PM- KNN 在所有实验中都超过了三种最先进的分类方法 。