The 1st International Conference on Computer Science and Engineering Technology Universitas Muria Kudus

Research Article

Clustering Sports News in Indonesian Using Modified K-Medoid Method

Download572 downloads
  • @INPROCEEDINGS{10.4108/eai.24-10-2018.2280569,
        author={Yoga Dwitya Pramudita and Sigit Susanto Putro and Eka Malasari Rochman and Ach. Yasir Rofiqi and Achmad Jauhari and Ika Oktavia Suzanti and Aeri Rachmad},
        title={Clustering Sports News in Indonesian Using Modified K-Medoid Method},
        proceedings={The 1st International Conference on Computer Science and Engineering Technology Universitas Muria Kudus},
        publisher={EAI},
        proceedings_a={ICCSET},
        year={2018},
        month={11},
        keywords={sport news speak clustering k-medoid modified k-medoid},
        doi={10.4108/eai.24-10-2018.2280569}
    }
    
  • Yoga Dwitya Pramudita
    Sigit Susanto Putro
    Eka Malasari Rochman
    Ach. Yasir Rofiqi
    Achmad Jauhari
    Ika Oktavia Suzanti
    Aeri Rachmad
    Year: 2018
    Clustering Sports News in Indonesian Using Modified K-Medoid Method
    ICCSET
    EAI
    DOI: 10.4108/eai.24-10-2018.2280569
Yoga Dwitya Pramudita1,*, Sigit Susanto Putro1, Eka Malasari Rochman1, Ach. Yasir Rofiqi1, Achmad Jauhari1, Ika Oktavia Suzanti1, Aeri Rachmad1
  • 1: University of Trunojoyo
*Contact email: yoga@trunojoyo.ac.id

Abstract

Sports news is a topic with high internet access rankings in Indonesia. The number of documents that must be managed continues to increase. In this research, news documents are grouped using the k-Medoid method. Medoid amount is adjusted to the number of news topics. In the k-Medoid method the cluster initialization process greatly influences the cluster results. Medoid is taken randomly according to the number of clusters desired. If the randomization results in a high degree of similarity, the resulting cluster is not optimal. Modifications are made to the initialization process so that the cluster has a low level of similarity. The test results showed accuracy reached 0.584 with five clusters. Modification of the k-Medoid method by adding the cosine similarity method can increase the average accuracy value from 0.44 to 0.54.