Comparison of K-Means and K-Medoids Algorithms for Grouping Cocoa Production Areas

Abstract

Cocoa is one of the leading commodities from the plantation sector, even cocoa production is considered capable of increasing the country's foreign exchange. In Indonesia, especially South Sulawesi Province, it has a large cocoa production where almost all districts/cities in South Sulawesi produce cocoa. The purpose of this research is to group cocoa production areas in South Sulawesi Province. The algorithms used are K-Means and K-Medoids, in which K-Means group data by dividing it into several clusters based on the same characteristics. While the K-Medoids algorithm chooses real objects to represent the cluster. In this study, the two algorithms were compared using one dataset. The comparison is made by looking at the Davies-Bouldin Index (DBI) value on RapidMiner. Then the results obtained based on this study are grouping using the K-Means algorithm is more effective than using K-Medoids in grouping cocoa production areas in South Sulawesi Province. With the DBI values ​​obtained, K-Means and K-Medoids have DBI values ​​of 0.292 and 0.365, respectively.