Comparison analysis of K-Means and K-Medoid with Ecluidience Distance Algorithm, Chanberra Distance, and Chebyshev Distance for Big Data Clustering

Abstract

This study aims to analyze the comparison of object clustering results in big data using K-Means method and K-Medoid method. Combination testing using three algorithms namely Ecludiance Distance, Canberra Distance, and Chebyshev Distance respectively for both methods. The sample of the research is the data set of students of three classes with six variables with a total of 147,679 data. The test results found that K-Means method is more optimal in data clustering than K-Medoid method, both in Ecluid Distance, Chanberra Distance and Chebyshev Distance algorithms which in overall comparison of clustering process with 1: 110.7 rationality. The results also suggested not to useChanberra Distance algorithms for both K-Means and K-Medoid methods because the cluster quality index in the Davies Bouldin test was undefined (∞). The best level of accuracy and quality of cloning is to use K-Means with Chebyshev Distance method, which is to produce five clusters with 0.1 second processing time