Investigation of Corruption Case Articles Using Self Organizing Maps (SOM) Method

Abstract

Nowadays, there are so many corruption articles in the internet with so many topics. Pusat Teknologi Informasi dan Komunikasi Badan Pengkajian dan Penerapan Teknologi (PTIK BPPT) conducts a research to make a system to cluster the corruption articles with related topic based on the keywords without changing the topology of the articles and can be visualized to the two dimension maps with information of related article and share rate. Self Organizing Maps (SOM) is the clustering method that is being used in the research. SOM can cluster massive article with high dimension to lower dimension without changing its topology. In this research, tools for measuring the clusters are Vector Quantization Error (VQE), cohesion, and separation. Method of computation cohesion and separation that used in this research is prototype-based cohesion and separation that focus in computing on the centroid. Finally, the conclusion of this research is that SOM clusters the articles based on the similarity of the topic. Neighbouring clusters mean that they have resemblance topics. Clusters that far-set from the other clusters means that they have less or do not have resemblance each other.