Local Mean k-Nearest Neighbor Classifier with Gaussian-Kernel Distance Function

Abstract

KNN is a simple but efficient machine learning classification algorithm that has an irreplaceable place in industry. In this paper, we propose a kernel-based local mean k-nearest neighbor classifier algorithm (KLMKNN) that not only considers the k nearest samples from each class of the samples to be classified, but also introduces a kernel mapping approach to compute data points in a high-dimensional space as a way to extract better location information. To verify the feasibility and efficiency of our proposed algorithm, we compared it with related algorithms on several datasets, and the results showed that the accuracy of our proposed algorithm is 0.7538 in all datasets, which is an improvement compared to KNN (0.7249), KKNN (0.7215) and LMKNN (0.7445), which is a good proof that our proposed algorithm is effective.