Proceedings of the 1st International Conference on Management, Business, Applied Science, Engineering and Sustainability Development, ICMASES 2019, 9-10 February 2019, Malang, Indonesia

Research Article

Learning Vector Quantization with Local Mean Based to Determine K Value in the K-Nearest Neighbor Method

Download456 downloads
  • @INPROCEEDINGS{10.4108/eai.3-8-2019.2290750,
        author={M A Munir and E B Nababan and Tulus  Tulus},
        title={Learning Vector Quantization with Local Mean Based to Determine K Value in the K-Nearest Neighbor Method},
        proceedings={Proceedings of the 1st International Conference on Management, Business, Applied Science, Engineering and Sustainability Development, ICMASES 2019, 9-10 February 2019, Malang, Indonesia},
        publisher={EAI},
        proceedings_a={ICMASES},
        year={2020},
        month={1},
        keywords={data iris learning vector quantization k-nearest neighbour local mean k-nearest neighbour k-fold cross-validation},
        doi={10.4108/eai.3-8-2019.2290750}
    }
    
  • M A Munir
    E B Nababan
    Tulus Tulus
    Year: 2020
    Learning Vector Quantization with Local Mean Based to Determine K Value in the K-Nearest Neighbor Method
    ICMASES
    EAI
    DOI: 10.4108/eai.3-8-2019.2290750
M A Munir1,*, E B Nababan2, Tulus Tulus3
  • 1: Graduate School of Computer Science, Faculty of Computer Science and Information Technology Universitas Sumatera Utara, Medan, Indonesia
  • 2: Department of Computer Science, Faculty of Computer Science and Information Technology, Universitas Sumatera Utara, Medan, Indonesia
  • 3: Department of Mathematics, Universitas Sumatera Utara, Medan, Indonesia
*Contact email: oreym29@gmail.com

Abstract

classification is a process that explains and functions to distinguish data classes or concepts that aim to be able to predictions in classes of objects unknown to the label class. Many popular classification techniques, one of which is K-Nearest Neighbor (KNN). The K-NN algorithm functions to find the closest k neighbors and use the majority class. This study aims to determine the best k value by using Learning Vector Quantization as weight weights. Determination of the Local Mean Based test data class K-Nearest Neighbor uses the measurement of the closest distance to each local model of each data class. In processing Learning Vector Quantization, Cross-Validation and Local K-Fold in the K-Nearest Neighbor classification the lowest k = 4 was 72%, while the highest k value was 9 = 80%. And the highest k value is a good K value that is k = 9 for Iris Data.