Multi-label classification of Indonesian news topics using Pseudo Nearest Neighbor Rule
Reza Agung Pambudi, Adiwijaya Adiwijaya, Mohamad Syahrul Mubarok · Journal of Physics Conference Series · 2019
News is a form of text data that must be categorized to facilitate retrieval of information for the reader.One problem that arises when categorizing news is the many topics that news can discuss, which is known as a multi-label condition.To solve this problem, a system that can perform multi-label classification using a Pseudo Nearest Neighbor Rule (PNNR) algorithm-a variant of the k-Nearest Neighbor (k-NNR) algorithm-was developed in this study.This system yielded a cross-validation error of 0,1495, measured using the hamming loss method via Cosine proximity.From the experiment, it can be concluded that the performance of the PNNR algorithm is influenced by the type of proximity used and the number of nearest neighbors.