Comparison Of K-Means and K-Medoids Algorithm for Clustering Data UMKM in Pagar Alam City

Abstract

The aim of this research is clustering MSME data in Pagar Alam City using the K-Means and K-Medoids algorithms. This research is motivated by the lack of further management of MSME data collection, which can hinder the development and improvement of Pagar Alam City MSMEs. Meanwhile, this data is considered necessary for agencies to develop and improve Pagar Alam City MSMEs. Apart from agencies, this data is also useful for sub-districts, sub-districts and RT/RW to find out what interests, talents and potential the community has in what business fields. MSME data is processed using Rapid Miner and Python, the system development method in this research uses the Cross Industry Standard Process for Data Mining (CRISP-DM) method, where the stages include Business Understanding, Data Understanding, Data Preparation, Modeling, Evaluation, and Deployment. The test method uses the Davies-bouldin index, a DBI value that is close to 0 results in good clustering. The results of this research obtained 3 clusters. In 2020 K-Means C0= 1, C1= 3 and C2= 1 sub-district, K-Medoids C0= 1, C1= 1 and C2= 3 sub-district. In 2022 K-Means C0= 1, C1= 3 and C2= 1 sub-district, K-Medoids C0= 1, C1= 3 and C2= 1 sub-districts. The results of the 2020 sub-district DBI clustering calculation are DBI k-means = 0.134 and k-medoids = 0.523. In 2022 DBI k-means = 0.277 and k-medoids = 0.496. So it can be concluded that the K-Means algorithm in the case of grouping MSMEs in Pagar Alam City has better performance, because the DBI value is close to 0. From the results of the grouping it can help provide an overview for related parties in encouraging or providing assistance to sub-districts that are included in the low cluster.