Application of K-Mean Algorithm for Medicine Data Clustering in Puskesmas Rumbai

Through the government’s health insurance program, efforts are made to ensure the health of the community through Puskesmas or community clinics. One of the most important components in health is the availability of medicines. The availability of medicines should be well managed to ensure that the medicines needed by the community are always available in sufficient quantities. Clustering on Data mining can be used to analyze the use of medicines during this time at a Puskesmas to be used as one of considerations for the Puskesmas to submit the demand of medicines in the period to come. The results of this study are expected to classify the level of medicines used in the pharmacy of Puskesmas in Rumbai Bukit Pekanbaru.


Introduction
Medicines is one of the important components in terms of good health to either prevent, reduce, eliminate or cure a disease or disease symptoms. That is why medicines need to be managed properly, effectively and efficiently. Planning on the needs of medicines is important to ensure the availability and distribution of medicines with the types and quantities sufficient so that medicines can be obtained quickly at the appropriate place and time at agencies related to health services, be it hospital, health centre, and so forth. The planning on the need of medicines will affect the procurement, distribution and usage of medicines in health care settings.Puskesmas Rumbai Bukit Pekanbaru is one of the public health service centres located in Pekanbaru which is government institution related to public health service in which one of its function isto provide medicine service for institution of health care settings in Pekanbaru. Effective and efficient analysis of the need of medicines is needed to ensure the availability of medicines in the pharmacy in Pekanbaru.
Clustering on the need of medicines is expected to be one of the considerations to ensure the availability of medicines in the health service of Pekanbaru. Data clustering is one method in data mining that can be used to get data mapping to classify into smaller groups based on the similarity of characteristics they have (Perim, Wandekokem, &Varejão, 2008). With these clustering results the distribution of medicines in health care agencies can be grouped according to the need based on the medicine distribution data in the previous year and it can be used as a reference for drug planning for the next year. It is hoped that the availability of medicines for the next year can be more secure and able to meet demand for medicines from health agencies. One of the most well-known clustering methods among clustering algorithms is K-means (Patel & Mehta, 2011). The simplicity of this method makes the K-means algorithm applicable to various fields (K.Arai and A.R. Barakbah, 2007).

Phase stages in data mining
KDD is a nontrivial process of identifying the validity of data, potential, use, and ultimately yielding understandable data patterns. The stages in the stages of data mining are: In this study, the data used are those taken from the Usage Report and Medicine Demand Sheet (LPLPO) Puskesmas Rumbai Bukit 2015. Then, the data are clustered to obtain patterns of medicine needs for the community of Rumbai in Pekanbaru.

Clustering
Clustering refers to grouping notes, observations, or cases into similar classes. A cluster is a collection of notes that resembles each other and differs from records in other clusters. Clustering is different from the classification which has no target variable for clustering. Instead, the algorithm clustering looks for the entire set of data segments into a relatively homogeneous subgroup or group, in which the similarity of records in the cluster is maximized, and the similarity of records out of this cluster is minimized.
Examples of grouping tasks in business and research include:  Calculating the z-score = −

K-Means Algorithm
Clustering k -means algorithm [1] is a simple and effective algorithm for finding clusters in data with the following algorithms: 1) Determine the number of cluster 2) Determine the value of the location of the initial cluster.
3) Calculate the closest cluster center for each record 4) For each cluster k, calculate the centroid cluster and update Location of each cluster center 5) Repeat steps 3 through 5 until convergence or termination.
The k-means algorithm is known and widely used for the partitional method, which is to divide the set of data objects into a subset of non-overlapping clusters, so that each data object is exactly in one cluster

Results and Discussion
The source of data used in this study is data from LPLPO Puskesmas Rumbai Bukit 2014. The data used can be seen in table 1.