Comparing K-Means and Fuzzy C-Means Clustering (Case: Clustering of Provinces in Indonesia Based on the Indicator of the Health Service in 2015)

Health is one of the main factors to develop and found of human resources. The health status of a country is greatly influenced by health service indicators such as health facilities, health workers and health financing. However, these indicators have not spreaded evenly throughout Indonesia. Therefore, grouping of provinces in Indonesia needs to be done in order to see the provincial groups that do not have adequate health services. One of the analysis methods is K-Means, which keeps an objects into a specific cluster. This method is known as hard clustering. Another approach in clustering is based on a fuzzy sets theory which is known as fuzzy clustering. Each element has the probability to become a member of each group like Fuzzy C-Means (FCM). The purpose of this research was to compare the K-Means and FCM in clustering case of provinces in Indonesia based on the health service indicators in 2015. Data used were secondary data taken from health profil of Indonesia in 2015 sourced from the result of the National Sosio-Economic Survey (Susenas). The best clustering was chosen by minimizing ratio of standard deviations in groups and between groups. The results showed that provinces of Indonesia were divided into five cluster with different characteristics. The best clustering was given by the FCM method which had the smallest ratio that was 1.385.