Comparison of K-Medoids and Clara Algorithm in Poverty Clustering Analysis in Indonesia

Abstract

The Covid-19 pandemic entered Indonesia in March 2020, so the government imposed restrictions on people's movement in various regencies. The imposition of restrictions on people's movement will have an impact on the economy to the point of poverty. Poverty is influenced by several factors such as population, health, education, employment and economic factors. The poverty of a district/city in Indonesia is grouped to assist the government in alleviating poverty more efficiently. The process of grouping data in data mining is to group districts/cities in Indonesia based on factors that affect poverty with the K-Medoids and CLARA algorithms, then compare the two methods based on the average value of the ratio of the standard deviations. The variables used in this study consist of 4 variables, namely human development index (HDI), gross regional domestic product (GRDP), unemployment rate, and population density. The results of this study indicate that using the K-Medoids obtained 2 clusters, while using the CLARA algorithm obtained 3 clusters. Based on the results of grouping the two algorithms, the best algorithm was obtained using cluster validation, namely the CLARA algorithm because it has the average value of the ratio of the smallest standard deviation of 0.106.