Application of the clustering algorithm to the small and micro industrial companies for mapping regions with k-medoids

Abstract

Abstract The largest segment of national economic actors is the micro and small enterprises (MSEs). The research aims to analyze the number of micro and small companies in Indonesia using data mining techniques to map a cluster. The dataset used consists of three-four records of the central Statistical Bureau (SourceUrl: https:/www.bps.go.id/), composed of the micro and smaller companies in 2017-2019. The k-medoids method is the solution used in cluster mapping. The average number of small and micro enterprises for 2017-2019 is the attribute used. The numbers of clusters were determined with the Davies Bouldin Index method (DBI), where k = 2 is the best value (0.111). In accordance with the results of the cluster (k=2), the label is divided into two (high cluster (clt1) and low group) (clt2). The results of the calculation of k-medoids show that 90% of Indonesia’s area is within the low class. In the high cluster are only Central Java, West Java and East Java. Accuracy, accuracy, reminder and f-measurement parameters are 100% demonstrated in the cluster test results. This mapping can be one of the foundations on which to constantly increase the number of micro and small firms, since the role of MSEs in jobs is extremely important, since it is apparent that MSEs have continued to grow considerably.