Determination value k in k-nearest nieghbor with local mean euclidean And weight gini index

2nd Nommensen International Conference on Technology and Engineering

Saputra, M.E.^1 ; Mawengkang, H.^2 ; Nababan, E.B.^3

Stud. in Faculty of Computer Science and Information Technology, Universitas Sumatera Utara, Medan, Indonesia^1

Faculty of Computer Science and Mathematics, Universitas Sumatera Utara, Medan, Indonesia^2

Faculty of Computer Science and Information Technology, Medan, Indonesia^3

关键词: Data class;    Euclidean;    Gini Index;    K-nearest neighbors;    Local mean;    Model based approach;    Nearest neighbors;    Unbalanced data;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/420/1/012098/pdf
DOI  :  10.1088/1757-899X/420/1/012098

来源: IOP

【 摘 要 】

K-Nearest Neighbor is the algorithm that included into the category of algorithms supervaised learning is bound process to distinguish the classes that already exists. Nearest neighbor is calculated based on the value of k that determines how the nearest neighbor in consider on the distance class data for k-nearest neighbor based on the determination of the value of the k. The determination of the class test data local mean based k-nearest neighbor using the measurement of the distance closest to each using eucllidean distance from each class data. Model based approach the weight of Gini Index is in need to give the weight of each attributes to determine the value of k. Research In this time I get the results of k best at the thyroid data which is a type of unbalanced data to obtain the value of the k highest k=44 until k=46 with accuracy of the closest neighbors of 71,19% and the value of the k lowest is k=50 of 69,30%. Then the results of the value of the k=44 until k=46 become k best on the processing of this time. It can be concluded for the data class is not the same will result in class data become random repeatedly until the limit of the determination of the value of k as well as exceeding the value of the k highest.