OPTIMIZATION OF K-NEAREST NEIGHBOUR TO CATEGORIZE INDONESIAN'S NEWS ARTICLES

Text classification is the process of grouping documents based on similarity in categories. Some obstacles doing text are many words appeared text, and some come up with infrequent frequency (sparse words). The way to solve this problem conduct feature selection process. There several filter-based methods; Chi-Square, Information Gain, Genetic Algorithm, Particle Swarm Optimization (PSO). Aghdam's research shows that PSO best among those methods. This study examined optimize k-Nearest Neighbour (k-NN) algorithm's performance categorizing news articles. k-NN an algorithm simple easy implement. If we use appropriate features, then will be a reliable algorithm. used select keywords (term features), it continued classifying using k-NN. testing consists three stages. stages tuning parameter k-NN, PSO, measuring performance. aims determine number neighbours particles. Otherwise, compares without PSO. optimal 9, particles 50. showed 50% reduction terms. results 20 per cent better accuracy than Although PSO's did not always find conditions, method can produce accuracy. In way, work articles, especially Indonesian language articles